,
Solve Sæbø [ctb],
Stefan Schrunner [rev]| Title: | Multiblock Data Fusion in Statistics and Machine Learning |
|---|---|
| Description: | Functions and datasets to support Smilde, Næs and Liland (2021, ISBN: 978-1-119-60096-1) "Multiblock Data Fusion in Statistics and Machine Learning - Applications in the Natural and Life Sciences". This implements and imports a large collection of methods for multiblock data analysis with common interfaces, result- and plotting functions, several real data sets and six vignettes covering a range different applications. |
| Authors: | Kristian Hovde Liland [aut, cre]
,
Solve Sæbø [ctb],
Stefan Schrunner [rev] |
| Maintainer: | Kristian Hovde Liland <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 0.8.8.2 |
| Built: | 2024-10-16 06:04:44 UTC |
| Source: | https://github.com/khliland/multiblock |
This is a quite general and flexible implementation of ASCA.
asca(formula, data, subset, weights, na.action, family, pca.in = FALSE)asca(formula, data, subset, weights, na.action, family, pca.in = FALSE)
formula |
Model formula accepting a single response (block) and predictor names separated by + signs. |
data |
The data set to analyse. |
subset |
Subset of objects |
weights |
Optional object weights. |
na.action |
How to handle NAs (no action implemented). |
family |
Error distributions and link function for Generalized Linear Models. |
pca.in |
Compress response before ASCA (number of components). |
ASCA is a method which decomposes a multivariate response according to one or more design
variables. ANOVA is used to split variation into contributions from factors, and PCA is performed
on the corresponding least squares estimates, i.e., Y = X1 B1 + X2 B2 + ... + E = T1 P1' + T2 P2' + ... + E.
This version of ASCA encompasses variants of LiMM-PCA, generalized ASCA and covariates ASCA. It includes
confidence ellipsoids for the balanced fixed effect ASCA.
An asca object containing loadings, scores, explained variances, etc. The object has
associated plotting (asca_plots) and result (asca_results) functions.
Smilde, A., Jansen, J., Hoefsloot, H., Lamers,R., Van Der Greef, J., and Timmerman, M.(2005). ANOVA-Simultaneous Component Analysis (ASCA): A new tool for analyzing designed metabolomics data. Bioinformatics, 21(13), 3043–3048.
Liland, K.H., Smilde, A., Marini, F., and Næs,T. (2018). Confidence ellipsoids for ASCA models based on multivariate regression theory. Journal of Chemometrics, 32(e2990), 1–13.
Martin, M. and Govaerts, B. (2020). LiMM-PCA: Combining ASCA+ and linear mixed models to analyse high-dimensional designed data. Journal of Chemometrics, 34(6), e3232.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in asca_results and asca_plots, respectively.
# Load candies data data(candies) # Basic ASCA model with two factors mod <- asca(assessment ~ candy + assessor, data=candies) print(mod) # ASCA model with interaction mod <- asca(assessment ~ candy * assessor, data=candies) print(mod) # Result plotting for first factor loadingplot(mod, scatter=TRUE, labels="names") scoreplot(mod, ellipsoids = "confidence") # ASCA model with compressed response using 5 principal components mod.pca <- asca(assessment ~ candy + assessor, data=candies, pca.in=5) # Mixed Model ASCA, random assessor mod.mix <- asca(assessment ~ candy + (1|assessor), data=candies) scoreplot(mod.mix)# Load candies data data(candies) # Basic ASCA model with two factors mod <- asca(assessment ~ candy + assessor, data=candies) print(mod) # ASCA model with interaction mod <- asca(assessment ~ candy * assessor, data=candies) print(mod) # Result plotting for first factor loadingplot(mod, scatter=TRUE, labels="names") scoreplot(mod, ellipsoids = "confidence") # ASCA model with compressed response using 5 principal components mod.pca <- asca(assessment ~ candy + assessor, data=candies, pca.in=5) # Mixed Model ASCA, random assessor mod.mix <- asca(assessment ~ candy + (1|assessor), data=candies) scoreplot(mod.mix)
Various plotting procedures for asca objects.
## S3 method for class 'asca' loadingplot(object, factor = 1, comps = 1:2, ...) ## S3 method for class 'asca' scoreplot( object, factor = 1, comps = 1:2, pch.scores = 19, pch.projections = 1, gr.col = 1:nlevels(object$effects[[factor]]), ellipsoids, confidence, xlim, ylim, xlab, ylab, legendpos, ... )## S3 method for class 'asca' loadingplot(object, factor = 1, comps = 1:2, ...) ## S3 method for class 'asca' scoreplot( object, factor = 1, comps = 1:2, pch.scores = 19, pch.projections = 1, gr.col = 1:nlevels(object$effects[[factor]]), ellipsoids, confidence, xlim, ylim, xlab, ylab, legendpos, ... )
object |
|
factor |
|
comps |
|
... |
additional arguments to underlying methods. |
pch.scores |
|
pch.projections |
|
gr.col |
|
ellipsoids |
|
confidence |
|
xlim |
|
ylim |
|
xlab |
|
ylab |
|
legendpos |
|
Usage of the functions are shown using generics in the examples in asca.
Plot routines are available as
scoreplot.asca and loadingplot.asca.
The plotting routines have no return.
Smilde, A., Jansen, J., Hoefsloot, H., Lamers,R., Van Der Greef, J., and Timmerman, M.(2005). ANOVA-Simultaneous Component Analysis (ASCA): A new tool for analyzing designed metabolomics data. Bioinformatics, 21(13), 3043–3048.
Liland, K.H., Smilde, A., Marini, F., and Næs,T. (2018). Confidence ellipsoids for ASCA models based on multivariate regression theory. Journal of Chemometrics, 32(e2990), 1–13.
Martin, M. and Govaerts, B. (2020). LiMM-PCA: Combining ASCA+ and linear mixed models to analyse high-dimensional designed data. Journal of Chemometrics, 34(6), e3232.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results are found in asca_results.
Standard result computation and extraction functions for ASCA (asca).
## S3 method for class 'asca' print(x, ...) ## S3 method for class 'asca' summary(object, ...) ## S3 method for class 'summary.asca' print(x, digits = 2, ...) ## S3 method for class 'asca' loadings(object, factor = 1, ...) ## S3 method for class 'asca' scores(object, factor = 1, ...) projections(object, ...) ## S3 method for class 'asca' projections(object, factor = 1, ...)## S3 method for class 'asca' print(x, ...) ## S3 method for class 'asca' summary(object, ...) ## S3 method for class 'summary.asca' print(x, digits = 2, ...) ## S3 method for class 'asca' loadings(object, factor = 1, ...) ## S3 method for class 'asca' scores(object, factor = 1, ...) projections(object, ...) ## S3 method for class 'asca' projections(object, factor = 1, ...)
x |
|
... |
additional arguments to underlying methods. |
object |
|
digits |
|
factor |
|
Usage of the functions are shown using generics in the examples in asca.
Explained variances are available (block-wise and global) through blockexpl and print.rosaexpl.
Object printing and summary are available through:
print.asca and summary.asca.
Scores and loadings have their own extensions of scores() and loadings() through
scores.asca and loadings.asca. Special to ASCA is that scores are on a
factor level basis, while back-projected samples have their own function in projections.asca.
Returns depend on method used, e.g. projections.asca returns projected samples,
scores.asca return scores, while print and summary methods return the object invisibly.
Smilde, A., Jansen, J., Hoefsloot, H., Lamers,R., Van Der Greef, J., and Timmerman, M.(2005). ANOVA-Simultaneous Component Analysis (ASCA): A new tool for analyzing designed metabolomics data. Bioinformatics, 21(13), 3043–3048.
Liland, K.H., Smilde, A., Marini, F., and Næs,T. (2018). Confidence ellipsoids for ASCA models based on multivariate regression theory. Journal of Chemometrics, 32(e2990), 1–13.
Martin, M. and Govaerts, B. (2020). LiMM-PCA: Combining ASCA+ and linear mixed models to analyse high-dimensional designed data. Journal of Chemometrics, 34(6), e3232.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for plotting are found in asca_plots.
This documentation covers a range of single- and two-block methods. In particular:
PCA - Principal Component Analysis (pca)
PCR - Principal Component Regression (pcr)
PLSR - Partial Least Squares Regression (plsr)
CCA - Canonical Correlation Analysis (cca)
IFA - Interbattery Factor Analysis (ifa)
GSVD - Generalized SVD (gsvd)
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(potato) X <- potato$Chemical y <- potato$Sensory[,1,drop=FALSE] pca.pot <- pca(X, ncomp = 2) pcr.pot <- pcr(y ~ X, ncomp = 2) pls.pot <- plsr(y ~ X, ncomp = 2) cca.pot <- cca(potato[1:2]) ifa.pot <- ifa(potato[1:2]) gsvd.pot <- gsvd(lapply(potato[3:4], t))data(potato) X <- potato$Chemical y <- potato$Sensory[,1,drop=FALSE] pca.pot <- pca(X, ncomp = 2) pcr.pot <- pcr(y ~ X, ncomp = 2) pls.pot <- plsr(y ~ X, ncomp = 2) cca.pot <- cca(potato[1:2]) ifa.pot <- ifa(potato[1:2]) gsvd.pot <- gsvd(lapply(potato[3:4], t))
This is a convenience function for making data.frames that are easily
indexed on a block-wise basis.
block.data.frame(X, block_inds = NULL, to.matrix = TRUE)block.data.frame(X, block_inds = NULL, to.matrix = TRUE)
X |
Either a single |
block_inds |
Named |
to.matrix |
|
A data.frame which can be indexed block-wise.
# Random data M <- matrix(rnorm(200), nrow = 10) # .. with dimnames dimnames(M) <- list(LETTERS[1:10], as.character(1:20)) # A named list for indexing inds <- list(B1 = 1:10, B2 = 11:20) X <- block.data.frame(M, inds) str(X)# Random data M <- matrix(rnorm(200), nrow = 10) # .. with dimnames dimnames(M) <- list(LETTERS[1:10], as.character(1:20)) # A named list for indexing inds <- list(B1 = 1:10, B2 = 11:20) X <- block.data.frame(M, inds) str(X)
A dataset containing 9 sensory attributes for 5 candies assessed by 11 trained assessors.
data(candies)data(candies)
A data.frame having 165 rows and 3 variables:
Matrix of sensory attributes
Factor of assessors
Factor of candies
Luciano G, Næs T. Interpreting sensory data by combining principal component analysis and analysis of variance. Food Qual Prefer. 2009;20(3):167-175.
This is a wrapper for the stats::cancor function for computing CCA.
cca(X)cca(X)
X |
|
CCA is a method which maximises correlation between linear combinations of the columns of two blocks, i.e. max(cor(X1 x a, X2 x b)). This is done sequentially with deflation in between, such that a sequence of correlations and weight vectors a and b are associated with a pair of matrices.
multiblock object with associated with printing, scores, loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Hotelling, H. (1936) Relations between two sets of variates. Biometrika, 28, 321–377.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) X <- potato$Chemical cca.pot <- cca(potato[1:2])data(potato) X <- potato$Chemical cca.pot <- cca(potato[1:2])
This documentation covers a few complex methods. In particular:
L-PLS - Partial Least Squares in L configuration (lpls)
SO-PLS-PM - Sequential and Orthogonalised PLS Path Modeling (sopls_pm)
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
# L-PLS sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 lp <- lpls(X1,X2,X3) # exo-L-PLS# L-PLS sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 lp <- lpls(X1,X2,X3) # exo-L-PLS
Convenience function for creating a vector
of component names based on the dimensions the input object
(matrix or object having a score function).
compnames(object, comps, explvar = FALSE, ...)compnames(object, comps, explvar = FALSE, ...)
object |
An object fitted using the multiblock package. |
comps |
|
explvar |
|
... |
Unused |
This is a copy of compnames from the pls package to work with
multiblock objects.
A character vector of component names.
This is a wrapper for the DISCOsca function by Zhengguo Gu for computing DISCO.
disco(X, ncomp = 2, ...)disco(X, ncomp = 2, ...)
X |
|
ncomp |
|
... |
additional arguments (not used). |
DISCO is a restriction of SCA where Alternating Least Squares is used for estimation of loadings and scores. The SCA solution is rotated towards loadings (in sample linked mode) which are filled with zeros in a pattern resembling distinct, local and common components. When used in sample linked mode and only selecting distinct components, it shares a resemblance to SO-PLS, only in an unsupervised setting. Explained variances are computed as proportion of block variation explained by scores*loadings'.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Schouteden, M., Van Deun, K., Wilderjans, T. F., & Van Mechelen, I. (2014). Performing DISCO-SCA to search for distinctive and common information in linked data. Behavior research methods, 46(2), 576-587.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.disco <- disco(potList) plot(scores(pot.disco), labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.disco <- disco(potList) plot(scores(pot.disco), labels="names")
A DISCO-SCA procedure for identifying common and distinctive components. The code is adapted from the orphaned RegularizedSCA package by Zhengguo Gu.
DISCOsca(DATA, R, Jk)DISCOsca(DATA, R, Jk)
DATA |
A matrix, which contains the concatenated data with the same subjects from multiple blocks. Note that each row represents a subject. |
R |
Number of components (R>=2). |
Jk |
A vector containing number of variables in the concatenated data matrix. |
Trot_best |
Estimated component score matrix (i.e., T) |
Prot_best |
Estimated component loading matrix (i.e., P) |
comdist |
A matrix representing common distinctive components. (Rows are data blocks and columns are components.) 0 in the matrix indicating that the corresponding
component of that block is estimated to be zeros, and 1 indicates that (at least one component loading in) the corresponding component of that block is not zero.
Thus, if a column in the |
propExp_component |
Proportion of variance per component. |
Schouteden, M., Van Deun, K., Wilderjans, T. F., & Van Mechelen, I. (2014). Performing DISCO-SCA to search for distinctive and common information in linked data. Behavior research methods, 46(2), 576-587.
## Not run: DATA1 <- matrix(rnorm(50), nrow=5) DATA2 <- matrix(rnorm(100), nrow=5) DATA <- cbind(DATA1, DATA2) R <- 5 Jk <- c(10, 20) DISCOsca(DATA, R, Jk) ## End(Not run)## Not run: DATA1 <- matrix(rnorm(50), nrow=5) DATA2 <- matrix(rnorm(100), nrow=5) DATA <- cbind(DATA1, DATA2) R <- 5 Jk <- c(10, 20) DISCOsca(DATA, R, Jk) ## End(Not run)
Flexible dummy-coding allowing for all R's built-in types of contrasts and optional dropping of a factor level to reduce rank defficiency probability.
dummycode(Y, contrast = "contr.sum", drop = TRUE)dummycode(Y, contrast = "contr.sum", drop = TRUE)
Y |
|
contrast |
Contrast type, default = "contr.sum". |
drop |
|
matrix made by dummy-coding the input vector.
vec <- c("a","a","b","b","c","c") dummycode(vec)vec <- c("a","a","b","b","c","c") dummycode(vec)
Extraction and/or computation of explained variances for various
object classes in the multiblock package.
explvar(object)explvar(object)
object |
An object fitted using a method from the multiblock package |
A vector of component-wise explained variances for predictors.
data(potato) so <- sopls(Sensory ~ Chemical + Compression, data=potato, ncomp=c(10,10), max_comps=10) explvar(so)data(potato) so <- sopls(Sensory ~ Chemical + Compression, data=potato, ncomp=c(10,10), max_comps=10) explvar(so)
This function attempts to apply model.frame and extend the
result with columns of interactions.
extended.model.frame(formula, data, ..., sep = ".")extended.model.frame(formula, data, ..., sep = ".")
formula |
a model formula or terms object or an R object. |
data |
a data.frame, list or environment (see |
... |
further arguments to pass to |
sep |
separator in contraction of names for interactions (default = "."). |
A data.frame that includes everything a model.frame
does plus interaction terms.
dat <- data.frame(Y = c(1,2,3,4,5,6), X = factor(LETTERS[c(1,1,2,2,3,3)]), W = factor(letters[c(1,2,1,2,1,2)])) extended.model.frame(Y ~ X*W, dat)dat <- data.frame(Y = c(1,2,3,4,5,6), X = factor(LETTERS[c(1,1,2,2,3,3)]), W = factor(letters[c(1,2,1,2,1,2)])) extended.model.frame(Y ~ X*W, dat)
This is an interface to both SVD-based (default) and RGCCA-based GCA (wrapping the
RGCCA::rgcca function)
gca(X, ncomp = "max", svd = TRUE, tol = 10^-12, corrs = TRUE, ...)gca(X, ncomp = "max", svd = TRUE, tol = 10^-12, corrs = TRUE, ...)
X |
|
ncomp |
|
svd |
|
tol |
|
corrs |
|
... |
additional arguments for RGCCA approach. |
GCA is a generalisation of Canonical Correlation Analysis to handle three or more
blocks. There are several ways to generalise, and two of these are available through gca.
The default is an SVD based approach estimating a common subspace and measuring mean squared
correlation to this. An alternative approach is available through RGCCA. For the SVD based
approach, the ncomp parameter controls the block-wise decomposition while the following
the consensus decomposition is limited to the minimum number of components from the individual blocks.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results. blockCoef contains canonical coefficients, while
blockDecomp contains decompositions of each block.
Carroll, J. D. (1968). Generalization of canonical correlation analysis to three or more sets of variables. Proceedings of the American Psychological Association, pages 227-22.
Van der Burg, E. and Dijksterhuis, G. (1996). Generalised canonical analysis of individual sensory profiles and instrument data, Elsevier, pp. 221–258.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.gca <- gca(potList) plot(scores(pot.gca), labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.gca <- gca(potList) plot(scores(pot.gca), labels="names")
This is a wrapper for the FactoMineR::GPA function for computing GPA.
gpa(X, graph = FALSE, ...)gpa(X, graph = FALSE, ...)
X |
|
graph |
|
... |
additional arguments for RGCCA approach. |
GPA is a generalisation of Procrustes analysis, where one matrix is scaled and rotated to be as similar as possible to another one. Through the generalisation, individual scaling and rotation of each input matrix is performed against a common representation which is estimated in an iterative manner.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Gower, J. C. (1975). Generalized procrustes analysis. Psychometrika. 40: 33–51.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.gpa <- gpa(potList) plot(scores(pot.gpa), labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.gpa <- gpa(potList) plot(scores(pot.gpa), labels="names")
This is a wrapper for the geigen::gsvd function for computing GSVD.
gsvd(X)gsvd(X)
X |
|
GSVD is a generalisation of SVD to two variable-linked matrices where common loadings and block-wise scores are estimated.
multiblock object with associated with printing, scores, loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Van Loan, C. (1976) Generalizing the singular value decomposition. SIAM Journal on Numerical Analysis, 13, 76–83.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) X <- potato$Chemical gsvd.pot <- gsvd(lapply(potato[3:4], t))data(potato) X <- potato$Chemical gsvd.pot <- gsvd(lapply(potato[3:4], t))
This is a simple implementation for computing HOGSVD
hogsvd(X)hogsvd(X)
X |
|
HOGSVD is a generalisation of SVD to two or more blocks. It finds a common set of loadings across blocks and individual sets of scores per block.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Ponnapalli, S. P., Saunders, M. A., Van Loan, C. F., & Alter, O. (2011). A higher-order generalized singular value decomposition for comparison of global mRNA expression from multiple organisms. PloS one, 6(12), e28072.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.hogsvd <- hogsvd(candyList) scoreplot(can.hogsvd, block=1, labels="names")data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.hogsvd <- hogsvd(candyList) scoreplot(can.hogsvd, block=1, labels="names")
This is a wrapper for the RGCCA::rgcca function for computing HPCA.
hpca(X, ncomp = 2, scale = FALSE, verbose = FALSE, ...)hpca(X, ncomp = 2, scale = FALSE, verbose = FALSE, ...)
X |
|
ncomp |
|
scale |
|
verbose |
|
... |
additional arguments for RGCCA. |
HPCA is a hierarchical PCA analysis which combines two or more blocks into a two-level decomposition with block-wise loadings and scores and superlevel common loadings and scores. The method is closely related to the supervised method MB-PLS in structure.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Westerhuis, J.A., Kourti, T., and MacGregor,J.F. (1998). Analysis of multiblock and hierarchical PCA and PLS models. Journal of Chemometrics, 12, 301–321.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.hpca <- hpca(potList) plot(scores(pot.hpca), labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.hpca <- hpca(potList) plot(scores(pot.hpca), labels="names")
This is a wrapper for the RGCCA::rgcca function for computing IFA.
ifa(X, ncomp = 1, scale = FALSE, verbose = FALSE, ...)ifa(X, ncomp = 1, scale = FALSE, verbose = FALSE, ...)
X |
|
ncomp |
|
scale |
|
verbose |
|
... |
additional arguments to |
IFA rotates two matrices to align one or more factors against each other, maximising correlations.
multiblock object with associated with printing, scores, loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Tucker, L. R. (1958). An inter-battery method of factor analysis. Psychometrika, 23(2), 111-136.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) X <- potato$Chemical ifa.pot <- ifa(potato[1:2])data(potato) X <- potato$Chemical ifa.pot <- ifa(potato[1:2])
This is a wrapper for the r.jive::jive function for computing JIVE.
jive(X, ...)jive(X, ...)
X |
|
... |
additional arguments for |
Jive performs a decomposition of the variation in two or more blocks into low-dimensional representations of individual and joint variation plus residual variation.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Lock, E., Hoadley, K., Marron, J., and Nobel, A. (2013) Joint and individual variation explained (JIVE) for integrated analysis of multiple data types. Ann Appl Stat, 7 (1), 523–542.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
# Too time consuming for testing data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.jive <- jive(candyList) summary(can.jive)# Too time consuming for testing data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.jive <- jive(candyList) summary(can.jive)
Simultaneous decomposition of three blocks connected in an L pattern.
lpls( X1, X2, X3, ncomp = 2, doublecenter = TRUE, scale = c(FALSE, FALSE, FALSE), type = c("exo"), impute = FALSE, niter = 25, subsetX2 = NULL, subsetX3 = NULL, ... )lpls( X1, X2, X3, ncomp = 2, doublecenter = TRUE, scale = c(FALSE, FALSE, FALSE), type = c("exo"), impute = FALSE, niter = 25, subsetX2 = NULL, subsetX3 = NULL, ... )
X1 |
|
X2 |
|
X3 |
|
ncomp |
number of L-PLS components |
doublecenter |
|
scale |
|
type |
|
impute |
|
niter |
|
subsetX2 |
|
subsetX3 |
|
... |
Additional arguments, not used. |
Two versions of L-PLS are available: exo- and endo-L-PLS which assume an outward or inward relationship between the main block X1 and the two other blocks X2 and X3.
The exo_ort algorithm returns orthogonal scores and should be chosen for visual
exploration in correlation loading plots. If exo-L-PLS with prediction is the main purpose
of the model then the non-orthogonal exo type L-PLS should be chosen for which the
predict function has prediction implemented.
An object of type lpls and multiblock containing all results from the L-PLS
analysis. The object type lpls is associated with functions for correlation loading plots,
prediction and cross-validation. The type multiblock is associated with the default functions
for result presentation (multiblock_results) and plotting (multiblock_plots).
Solve Sæbø (adapted by Kristian Hovde Liland)
Martens, H., Anderssen, E., Flatberg, A.,Gidskehaug, L.H., Høy, M., Westad, F.,Thybo, A., and Martens, M. (2005). Regression of a data matrix on descriptors of both its rows and of its columns via latent variables: L-PLSR. Computational Statistics & Data Analysis, 48(1), 103 – 123.
Sæbø, S., Almøy, T., Flatberg, A., Aastveit, A.H., and Martens, H. (2008). LPLS-regression: a method for prediction and classification under the influence of background information on predictor variables. Chemometrics and Intelligent Laboratory Systems, 91, 121–132.
Sæbø, S., Martens, M. and Martens H. (2010) Three-block data modeling by endo- and exo-LPLS regression. In Handbook of Partial Least Squares: Concepts, Methods and Applications. Esposito Vinzi, V.; Chin, W.W.; Henseler, J.; Wang, H. (Eds.). Springer.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Functions for computation and extraction of results and plotting are found in lpls_results.
# Simulate data set sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 lp <- lpls(X1,X2,X3) # exo-L-PLS# Simulate data set sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 lp <- lpls(X1,X2,X3) # exo-L-PLS
lpls)Correlation loading plot, prediction and cross-validation for L-PLS
models with class lpls.
## S3 method for class 'lpls' plot( x, comps = c(1, 2), doplot = c(TRUE, TRUE, TRUE), level = c(2, 2, 2), arrow = c(1, 0, 1), xlim = c(-1, 1), ylim = c(-1, 1), samplecol = 4, pathcol = 2, varcol = "grey70", varsize = 1, sampleindex = 1:dim(x$corloadings$R22)[1], pathindex = 1:dim(x$corloadings$R3)[1], varindex = 1:dim(x$corloadings$R21)[1], ... ) ## S3 method for class 'lpls' predict( object, X1new = NULL, X2new = NULL, X3new = NULL, exo.direction = c("X2", "X3"), ... ) lplsCV(object, segments1 = NULL, segments2 = NULL, trace = TRUE)## S3 method for class 'lpls' plot( x, comps = c(1, 2), doplot = c(TRUE, TRUE, TRUE), level = c(2, 2, 2), arrow = c(1, 0, 1), xlim = c(-1, 1), ylim = c(-1, 1), samplecol = 4, pathcol = 2, varcol = "grey70", varsize = 1, sampleindex = 1:dim(x$corloadings$R22)[1], pathindex = 1:dim(x$corloadings$R3)[1], varindex = 1:dim(x$corloadings$R21)[1], ... ) ## S3 method for class 'lpls' predict( object, X1new = NULL, X2new = NULL, X3new = NULL, exo.direction = c("X2", "X3"), ... ) lplsCV(object, segments1 = NULL, segments2 = NULL, trace = TRUE)
x |
|
comps |
|
doplot |
|
level |
|
arrow |
|
xlim |
|
ylim |
|
samplecol |
|
pathcol |
|
varcol |
|
varsize |
|
sampleindex |
|
pathindex |
|
varindex |
|
... |
Not implemented. |
object |
|
X1new |
|
X2new |
|
X3new |
|
exo.direction |
|
segments1 |
|
segments2 |
|
trace |
|
Nothing is return for plotting (plot.lpls), predicted values are returned for predictions (predict.lpls)
and cross-validation metrics are returned for for cross-validation (lplsCV).
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
# Simulate data set sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 # exo-L-PLS: lp.exo <- lpls(X1,X2,X3, ncomp = 2) # Predict X1 pred.exo.X2 <- predict(lp.exo, X1new = X1, exo.direction = "X2") # Predict X3 pred.exo.X2 <- predict(lp.exo, X1new = X1, exo.direction = "X3") # endo-L-PLS: lp.endo <- lpls(X1,X2,X3, ncomp = 2, type = "endo") # Predict X1 from X2 and X3 (in this case fitted values): pred.endo.X1 <- predict(lp.endo, X2new = X2, X3new = X3) # LOO cross-validation horizontally lp.cv1 <- lplsCV(lp.exo, segments1 = as.list(1:dim(X1)[1])) # LOO cross-validation vertically lp.cv2 <- lplsCV(lp.exo, segments2 = as.list(1:dim(X1)[2])) # Three-fold CV, horizontal lp.cv3 <- lplsCV(lp.exo, segments1 = as.list(1:10, 11:20, 21:30))# Simulate data set sim <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) X1 <- sim$X1; X2 <- sim$X2; X3 <- sim$X3 # exo-L-PLS: lp.exo <- lpls(X1,X2,X3, ncomp = 2) # Predict X1 pred.exo.X2 <- predict(lp.exo, X1new = X1, exo.direction = "X2") # Predict X3 pred.exo.X2 <- predict(lp.exo, X1new = X1, exo.direction = "X3") # endo-L-PLS: lp.endo <- lpls(X1,X2,X3, ncomp = 2, type = "endo") # Predict X1 from X2 and X3 (in this case fitted values): pred.endo.X1 <- predict(lp.endo, X2new = X2, X3new = X3) # LOO cross-validation horizontally lp.cv1 <- lplsCV(lp.exo, segments1 = as.list(1:dim(X1)[1])) # LOO cross-validation vertically lp.cv2 <- lplsCV(lp.exo, segments2 = as.list(1:dim(X1)[2])) # Three-fold CV, horizontal lp.cv3 <- lplsCV(lp.exo, segments1 = as.list(1:10, 11:20, 21:30))
Three data blocks are simulated to express covariance in an exo-L-PLS direction (see lpls.
Dimensionality and number of underlying components can be controlled.
lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2)lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2)
I |
|
N |
|
J |
|
K |
|
ncomp |
|
A list of three matrices with dimensions matching in an L-shape.
Solve Sæbø (adapted by Kristian Hovde Liland)
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
lp <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) names(lp)lp <- lplsData(I = 30, N = 20, J = 5, K = 6, ncomp = 2) names(lp)
Måge plot for SO-PLS (sopls) cross-validation visualisation.
maage( object, expl_var = TRUE, pure.trace = FALSE, pch = 20, xlab = "# components", ylab = ifelse(expl_var, "Explained variance (%)", "RMSECV"), xlim = NULL, ylim = NULL, cex.text = 0.8, ... ) maageSeq( object, compSeq = TRUE, expl_var = TRUE, pch = 20, xlab = "# components", ylab = ifelse(expl_var, "Explained variance (%)", "RMSECV"), xlim = NULL, ylim = NULL, cex.text = 0.8, col = "gray", col.block = c("red", "blue", "darkgreen", "purple", "black", "red", "blue", "darkgreen"), ... )maage( object, expl_var = TRUE, pure.trace = FALSE, pch = 20, xlab = "# components", ylab = ifelse(expl_var, "Explained variance (%)", "RMSECV"), xlim = NULL, ylim = NULL, cex.text = 0.8, ... ) maageSeq( object, compSeq = TRUE, expl_var = TRUE, pch = 20, xlab = "# components", ylab = ifelse(expl_var, "Explained variance (%)", "RMSECV"), xlim = NULL, ylim = NULL, cex.text = 0.8, col = "gray", col.block = c("red", "blue", "darkgreen", "purple", "black", "red", "blue", "darkgreen"), ... )
object |
An SO-PLS model ( |
expl_var |
Logical indicating if explained variance (default) or RMSECV should be displayed. |
pure.trace |
Logical indicating if single block solutions should be traced in the plot. |
pch |
Scalar or symbol giving plot symbol. |
xlab |
Label for x-axis. |
ylab |
Label for y-axis. |
xlim |
Plot limits for x-axis (numeric vector). |
ylim |
Plot limits for y-axis (numeric vector). |
cex.text |
Text scaling (scalar) for better readability of plots. |
... |
Additional arguments to |
compSeq |
Integer vector giving the sequence of previous components chosen for |
col |
Line colour in plot. |
col.block |
Line colours for blocks (default = c('red','blue','darkgreen','purple','black')) |
This function can either be used
for global optimisation across blocks or sequential optimisation, using maageSeq.
The examples below show typical usage.
The maage plot has no return.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(wine) ncomp <- unlist(lapply(wine, ncol))[-5] so.wine <- sopls(`Global quality` ~ ., data=wine, ncomp=ncomp, max_comps=10, validation="CV", segments=10) maage(so.wine) # Sequential search for optimal number of components per block old.par <- par(mfrow=c(2,2), mar=c(3,3,0.5,1), mgp=c(2,0.7,0)) maageSeq(so.wine) maageSeq(so.wine, 2) maageSeq(so.wine, c(2,1)) maageSeq(so.wine, c(2,1,1)) par(old.par)data(wine) ncomp <- unlist(lapply(wine, ncol))[-5] so.wine <- sopls(`Global quality` ~ ., data=wine, ncomp=ncomp, max_comps=10, validation="CV", segments=10) maage(so.wine) # Sequential search for optimal number of components per block old.par <- par(mfrow=c(2,2), mar=c(3,3,0.5,1), mgp=c(2,0.7,0)) maageSeq(so.wine) maageSeq(so.wine, 2) maageSeq(so.wine, c(2,1)) maageSeq(so.wine, c(2,1,1)) par(old.par)
A function computing MB-PLS scores, loadings, etc. on the super-level and block-level.
mbpls( formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, scale = FALSE, blockScale = c("sqrtnvar", "ssq", "none"), ... )mbpls( formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, scale = FALSE, blockScale = c("sqrtnvar", "ssq", "none"), ... )
formula |
Model formula accepting a single response (block) and predictor block names separated by + signs. |
data |
The data set to analyse. |
subset |
Expression for subsetting the data before modelling. |
na.action |
How to handle NAs (no action implemented). |
X |
|
Y |
|
ncomp |
|
scale |
|
blockScale |
Either a |
... |
additional arguments to pls::plsr. |
MB-PLS is the prototypical component based supervised multiblock method.
It was originally formulated as a two-level method with a block-level and a super-level,
but it was later discovered that it could be expressed as an ordinary PLS on concatenated
weighted X blocks followed by a simple loop for calculating block-level loading weights,
loadings and scores. This implementation uses the plsr function on the
scaled input blocks (1/sqrt(ncol)) enabling all summaries and plots from the pls
package.
Block weighting is performed after scaling all variables and is by default
"sqrtnvar": 1/sqrt(ncol(X[[i]])) in each block. Alternatives
are "ssq": 1/norm(X[[i]], "F")^2 and "none": 1/1. Finally, if
a numeric vector is supplied, it will be used to scale the blocks
after "ssq" scaling, i.e., Z[[i]] = X[[i]] / norm(X[[i]], "F")^2 * blockScale[i].
multiblock, mvr object with super-scores, super-loadings, block-scores and block-loading, and the underlying
mvr (PLS) object for the super model, with all its result and plot possibilities. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Wangen, L.E. and Kowalski, B.R. (1988). A multiblock partial least squares algorithm for investigating complex chemical systems. Journal of Chemometrics, 3, 3–20.
Westerhuis, J.A., Kourti, T., and MacGregor,J.F. (1998). Analysis of multiblock and hierarchical PCA and PLS models. Journal of Chemometrics, 12, 301–321.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(potato) # Formula interface mb <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5) # ... or X and Y mb.XY <- mbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5) identical(mb$scores, mb.XY$scores) print(mb) scoreplot(mb, labels="names") # Exploiting mvr object structure from pls package # Block scaling with emphasis on first block mbs <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, blockScale = c(10, 1)) scoreplot(mbs, labels="names") # Exploiting mvr object structure from pls packagedata(potato) # Formula interface mb <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5) # ... or X and Y mb.XY <- mbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5) identical(mb$scores, mb.XY$scores) print(mb) scoreplot(mb, labels="names") # Exploiting mvr object structure from pls package # Block scaling with emphasis on first block mbs <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, blockScale = c(10, 1)) scoreplot(mbs, labels="names") # Exploiting mvr object structure from pls package
This is a wrapper for the ade4::mbpcaiv function for computing mbRDA.
mbrda(formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, ...)mbrda(formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, ...)
formula |
Model formula accepting a single response (block) and predictor block names separated by + signs. |
data |
The data set to analyse. |
subset |
Expression for subsetting the data before modelling. |
na.action |
How to handle NAs (no action implemented). |
X |
|
Y |
|
ncomp |
|
... |
additional arguments to ade4::mbpcaiv. |
mbRDA is a multiblock formulation of Redundancy (Data) Analysis. RDA is theoretically
between PLS and GCA. Like GCA, RDA does not consider correlations within X, but like
PLS it does consider correlations within Y. RDA can also be viewed as a PCR of Y constrained to
have scores that are also linear combinations of X. If the adegraphics package is attached,
a nice overview can be plotted as plot(mbr$mbpcaiv) following the example below.
multiblock, mvr object with scores, block-scores and block-loading. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Bougeard, S., Qannari, E.M., Lupo, C., andHanafi, M. (2011). From Multiblock Partial Least Squares to Multiblock Redundancy Analysis. A Continuum Approach. Informatica, 22(1), 11–26.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
# Convert data.frame with AsIs objects to list of matrices data(potato) potatoList <- lapply(potato, unclass) mbr <- mbrda(Sensory ~ Chemical + Compression, data = potatoList, ncomp = 10) mbr.XY <- mbrda(X = potatoList[c('Chemical','Compression')], Y = potatoList[['Sensory']], ncomp = 10) print(mbr) scoreplot(mbr) # Exploiting mvr object structure from pls package# Convert data.frame with AsIs objects to list of matrices data(potato) potatoList <- lapply(potato, unclass) mbr <- mbrda(Sensory ~ Chemical + Compression, data = potatoList, ncomp = 10) mbr.XY <- mbrda(X = potatoList[c('Chemical','Compression')], Y = potatoList[['Sensory']], ncomp = 10) print(mbr) scoreplot(mbr) # Exploiting mvr object structure from pls package
This is a wrapper for the RGCCA::rgcca function for computing MCOA.
mcoa(X, ncomp = 2, scale = FALSE, verbose = FALSE, ...)mcoa(X, ncomp = 2, scale = FALSE, verbose = FALSE, ...)
X |
|
ncomp |
|
scale |
|
verbose |
|
... |
additional arguments for RGCCA. |
MCOA resembles GCA and MFA in that it creates a set of reference scores, for which each block's individual scores should correlate maximally too, but also the variance within each block should be taken into account. A single component solution is equivalent to a PCA on concatenated blocks scaled by the so called inverse inertia.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Le Roux; B. and H. Rouanet (2004). Geometric Data Analysis, From Correspondence Analysis to Structured Data Analysis. Dordrecht. Kluwer: p.180.
Greenacre, Michael and Blasius, Jörg (editors) (2006). Multiple Correspondence Analysis and Related Methods. London: Chapman & Hall/CRC.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.mcoa <- mcoa(potList) plot(scores(pot.mcoa), labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.mcoa <- mcoa(potList) plot(scores(pot.mcoa), labels="names")
Colour palette generation from matrix of RGB values
mcolors( n, colmatrix = matrix(c(0, 0, 1, 1, 1, 1, 1, 0, 0), 3, 3, byrow = TRUE) )mcolors( n, colmatrix = matrix(c(0, 0, 1, 1, 1, 1, 1, 0, 0), 3, 3, byrow = TRUE) )
n |
Integer number of colorus to produce. |
colmatrix |
A numeric |
A vector of n colours in hexadecimal RGB format.
mcolors(5)mcolors(5)
This is a wrapper for the FactoMineR::MFA function for computing MFA.
mfa(X, type = rep("c", length(X)), graph = FALSE, ...)mfa(X, type = rep("c", length(X)), graph = FALSE, ...)
X |
|
type |
|
graph |
|
... |
additional arguments for RGCCA approach. |
MFA is a methods typically used to compare several equally sized matrices. It is often used in sensory analyses, where matrices consist of sensory characteristics and products, and each assessor generates one matrix each. In its basic form, MFA scales all matrices by their largest eigenvalue, concatenates them and performs PCA on the result. There are several possibilities for plots and inspections of the model, handling of categorical and continuous inputs etc. connected to MFA.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Pagès, J. (2005). Collection and analysis of perceived product inter-distances using multiple factor analysis: Application to the study of 10 white wines from the Loire valley. Food Quality and Preference, 16(7), 642–649.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.mfa <- mfa(potList) if(interactive()){ plot(pot.mfa$MFA) }data(potato) potList <- as.list(potato[c(1,2,9)]) pot.mfa <- mfa(potList) if(interactive()){ plot(pot.mfa$MFA) }
Plotting procedures for multiblock objects.
## S3 method for class 'multiblock' scoreplot( object, comps = 1:2, block = 0, labels, identify = FALSE, type = "p", xlab, ylab, main, ... ) ## S3 method for class 'multiblock' loadingplot( object, comps = 1:2, block = 0, scatter = TRUE, labels, identify = FALSE, type, lty, lwd = NULL, pch, cex = NULL, col, legendpos, xlab, ylab, main, pretty.xlabels = TRUE, xlim, ... ) loadingweightplot(object, main = "Loading weights", ...) ## S3 method for class 'multiblock' biplot( x, block = 0, comps = 1:2, which = c("x", "y", "scores", "loadings"), var.axes = FALSE, xlabs, ylabs, main, ... ) corrplot(object, ...) ## Default S3 method: corrplot(object, ...) ## S3 method for class 'mvr' corrplot(object, ...) ## S3 method for class 'multiblock' corrplot( object, comps = 1:2, labels = TRUE, col = 1:5, plotx = TRUE, ploty = TRUE, blockScores = FALSE, ... )## S3 method for class 'multiblock' scoreplot( object, comps = 1:2, block = 0, labels, identify = FALSE, type = "p", xlab, ylab, main, ... ) ## S3 method for class 'multiblock' loadingplot( object, comps = 1:2, block = 0, scatter = TRUE, labels, identify = FALSE, type, lty, lwd = NULL, pch, cex = NULL, col, legendpos, xlab, ylab, main, pretty.xlabels = TRUE, xlim, ... ) loadingweightplot(object, main = "Loading weights", ...) ## S3 method for class 'multiblock' biplot( x, block = 0, comps = 1:2, which = c("x", "y", "scores", "loadings"), var.axes = FALSE, xlabs, ylabs, main, ... ) corrplot(object, ...) ## Default S3 method: corrplot(object, ...) ## S3 method for class 'mvr' corrplot(object, ...) ## S3 method for class 'multiblock' corrplot( object, comps = 1:2, labels = TRUE, col = 1:5, plotx = TRUE, ploty = TRUE, blockScores = FALSE, ... )
object |
|
comps |
|
block |
|
labels |
|
identify |
|
type |
|
xlab |
|
ylab |
|
main |
|
... |
Not implemented. |
scatter |
|
lty |
Vector of line type specifications (see |
lwd |
|
pch |
Vector of point specifications (see |
cex |
|
col |
|
legendpos |
|
pretty.xlabels |
|
xlim |
|
x |
|
which |
|
var.axes |
|
xlabs |
|
ylabs |
|
plotx |
|
ploty |
|
blockScores |
|
Plot functions for scores, loadings and loading.weights based
on the functions found in the pls package.
These plotting routines only generate plots and return no values.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results are found in multiblock_results.
data(wine) sc <- sca(wine[c('Smell at rest', 'View', 'Smell after shaking')], ncomp = 4) loadingplot(sc, block = 1, labels = "names", scatter = TRUE) scoreplot(sc, labels = "names") corrplot(sc) data(potato) so <- sopls(Sensory ~ NIRraw + Chemical + Compression, data=potato, ncomp = c(2,2,2), max_comps = 6, validation = "CV", segments = 10) scoreplot(so, ncomp = c(2,1), block = 3, labels = "names") corrplot(pcp(so, ncomp = c(2,2,2)))data(wine) sc <- sca(wine[c('Smell at rest', 'View', 'Smell after shaking')], ncomp = 4) loadingplot(sc, block = 1, labels = "names", scatter = TRUE) scoreplot(sc, labels = "names") corrplot(sc) data(potato) so <- sopls(Sensory ~ NIRraw + Chemical + Compression, data=potato, ncomp = c(2,2,2), max_comps = 6, validation = "CV", segments = 10) scoreplot(so, ncomp = c(2,1), block = 3, labels = "names") corrplot(pcp(so, ncomp = c(2,2,2)))
Standard result computation and extraction functions for multiblock objects.
## S3 method for class 'multiblock' scores(object, block = 0, ...) ## S3 method for class 'multiblock' loadings(object, block = 0, ...) ## S3 method for class 'multiblock' print(x, ...) ## S3 method for class 'multiblock' summary(object, ...)## S3 method for class 'multiblock' scores(object, block = 0, ...) ## S3 method for class 'multiblock' loadings(object, block = 0, ...) ## S3 method for class 'multiblock' print(x, ...) ## S3 method for class 'multiblock' summary(object, ...)
object |
|
block |
|
... |
Not implemented. |
x |
|
Usage of the functions are shown using generics in the examples below.
Object printing and summary are available through:
print.multiblock and summary.multiblock.
Scores and loadings have their own extensions of scores() and loadings() throught
scores.multiblock and loadings.multiblock.
Scores or loadings are returned by scores.multiblock and loadings.multiblock, while print and summary methods invisibly returns the object.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for plotting are found in multiblock_plots, respectively.
data(wine) sc <- sca(wine[c('Smell at rest', 'View', 'Smell after shaking')], ncomp = 4) print(sc) summary(sc) head(loadings(sc, block = 1)) head(scores(sc))data(wine) sc <- sca(wine[c('Smell at rest', 'View', 'Smell after shaking')], ncomp = 4) print(sc) summary(sc) head(loadings(sc, block = 1)) head(scores(sc))
Functions to estimate the mean squared error of prediction (MSEP), root mean
squared error of prediction (RMSEP) and (A.K.A. coefficient of
multiple determination) for a fitted MB-PLS models. Test-set,
cross-validation and calibration-set estimates are implemented.
## S3 method for class 'mbpls' R2( object, estimate, newdata, ncomp = 1:object$ncomp, comps, intercept = TRUE, se = FALSE, ... ) ## S3 method for class 'mbpls' MSEP( object, estimate, newdata, ncomp = 1:object$ncomp, comps, intercept = TRUE, se = FALSE, ... ) ## S3 method for class 'mbpls' RMSEP(object, ...)## S3 method for class 'mbpls' R2( object, estimate, newdata, ncomp = 1:object$ncomp, comps, intercept = TRUE, se = FALSE, ... ) ## S3 method for class 'mbpls' MSEP( object, estimate, newdata, ncomp = 1:object$ncomp, comps, intercept = TRUE, se = FALSE, ... ) ## S3 method for class 'mbpls' RMSEP(object, ...)
object |
an |
estimate |
a character vector. Which estimators to use. Should be a
subset of |
newdata |
a data frame with test set data. |
ncomp, comps
|
a vector of positive integers. The components or number of components to use. See below. |
intercept |
logical. Whether estimates for a model with zero components should be returned as well. |
se |
logical. Whether estimated standard errors of the estimates should be calculated. Not implemented yet. |
... |
further arguments sent to underlying functions or (for
|
RMSEP simply calls MSEP and takes the square root of the
estimates. It therefore accepts the same arguments as MSEP.
Several estimators can be used. "train" is the training or
calibration data estimate, also called (R)MSEC. For R2, this is the
unadjusted . It is overoptimistic and should not be used for
assessing models. "CV" is the cross-validation estimate, and
"adjCV" (for RMSEP and MSEP) is the bias-corrected
cross-validation estimate. They can only be calculated if the model has
been cross-validated. Finally, "test" is the test set estimate,
using newdata as test set.
Which estimators to use is decided as follows (see below for
pls:mvrValstats). If estimate is not specified, the test set
estimate is returned if newdata is specified, otherwise the CV and
adjusted CV (for RMSEP and MSEP) estimates if the model has
been cross-validated, otherwise the training data estimate. If
estimate is "all", all possible estimates are calculated.
Otherwise, the specified estimates are calculated.
Several model sizes can also be specified. If comps is missing (or
is NULL), length(ncomp) models are used, with ncomp[1]
components, ..., ncomp[length(ncomp)] components. Otherwise, a
single model with the components comps[1], ...,
comps[length(comps)] is used. If intercept is TRUE, a
model with zero components is also used (in addition to the above).
The values returned by "R2" are calculated as , where is the (corrected) total sum of squares of the
response, and is the sum of squared errors for either the fitted
values (i.e., the residual sum of squares), test set predictions or
cross-validated predictions (i.e., the ). For estimate =
"train", this is equivalent to the squared correlation between the fitted
values and the response. For estimate = "train", the estimate is
often called the prediction .
mvrValstats is a utility function that calculates the statistics
needed by MSEP and R2. It is not intended to be used
interactively. It accepts the same arguments as MSEP and R2.
However, the estimate argument must be specified explicitly: no
partial matching and no automatic choice is made. The function simply
calculates the types of estimates it knows, and leaves the other untouched.
mvrValstats returns a list with components
three-dimensional array of SSE values. The first dimension is the different estimators, the second is the response variables and the third is the models.
matrix of SST values. The first dimension is the different estimators and the second is the response variables.
a numeric vector giving the number of objects used for each estimator.
the components specified, with 0 prepended
if intercept is TRUE.
TRUE if
comps was NULL or not specified.
The other functions return an object of class "mvrVal", with
components
three-dimensional array of estimates. The first dimension is the different estimators, the second is the response variables and the third is the models.
"MSEP",
"RMSEP" or "R2".
the components specified, with
0 prepended if intercept is TRUE.
TRUE if comps was NULL or not
specified.
the function call
Kristian Hovde Liland
Mevik, B.-H., Cederkvist, H. R. (2004) Mean Squared Error of Prediction (MSEP) Estimates for Principal Component Regression (PCR) and Partial Least Squares Regression (PLSR). Journal of Chemometrics, 18(9), 422–429.
data(oliveoil, package = "pls") mod <- pls::plsr(sensory ~ chemical, ncomp = 4, data = oliveoil, validation = "LOO") RMSEP(mod) ## Not run: plot(R2(mod))data(oliveoil, package = "pls") mod <- pls::plsr(sensory ~ chemical, ncomp = 4, data = oliveoil, validation = "LOO") RMSEP(mod) ## Not run: plot(R2(mod))
This is a wrapper for the pls::PCR function for computing PCA.
pca(X, scale = FALSE, ncomp = 1, ...)pca(X, scale = FALSE, ncomp = 1, ...)
X |
|
scale |
|
ncomp |
|
... |
additional arguments to |
PCA is a method for decomposing a matrix into subspace components with sample scores and variable loadings. It can be formulated in various ways, but the standard formulation uses singular value decomposition to create scores and loadings. PCA is guaranteed to be the optimal way of extracting orthogonal subspaces from a matrix with regard to the amount of explained variance per component.
multiblock object with scores, loadings, mean X values and explained variances. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Pearson, K. (1901) On lines and planes of closest fit to points in space. Philosophical Magazine, 2, 559–572.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) X <- potato$Chemical pca.pot <- pca(X, ncomp = 2)data(potato) X <- potato$Chemical pca.pot <- pca(X, ncomp = 2)
PCA-GCA is a methods which aims at estimating subspaces of common, local and distinct variation from two or more blocks.
pcagca( X, commons = 2, auto = TRUE, auto.par = list(explVarLim = 40, rLim = 0.8), manual.par = list(ncomp = 0, ncommon = 0), tol = 10^-12 )pcagca( X, commons = 2, auto = TRUE, auto.par = list(explVarLim = 40, rLim = 0.8), manual.par = list(ncomp = 0, ncommon = 0), tol = 10^-12 )
X |
|
commons |
|
auto |
|
auto.par |
|
manual.par |
|
tol |
|
The name PCA-GCA comes from the process of first applying PCA to each block, then using GCA to estimate local and common components, and finally orthogonalising the block-wise scores on the local/common ones and re-estimating these to obtain distinct components. The procedure is highly similar to the supervised method PO-PLS, where the PCA steps are exchanged with PLS.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results. Distinct components are marked as 'D(x), Comp c' for block x and component c
while local and common components are marked as "C(x1, x2), Comp c", where x1 and x2 (and more) are block numbers.
Smilde, A., Måge, I., Naes, T., Hankemeier, T.,Lips, M., Kiers, H., Acar, E., and Bro, R.(2017). Common and distinct components in data fusion. Journal of Chemometrics, 31(7), e2900.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) potList <- as.list(potato[c(1,2,9)]) pot.pcagca <- pcagca(potList) # Show origin and type of all components lapply(pot.pcagca$blockScores,colnames) # Basic multiblock plot plot(scores(pot.pcagca, block=2), comps=1, labels="names")data(potato) potList <- as.list(potato[c(1,2,9)]) pot.pcagca <- pcagca(potList) # Show origin and type of all components lapply(pot.pcagca$blockScores,colnames) # Basic multiblock plot plot(scores(pot.pcagca, block=2), comps=1, labels="names")
This is a basic implementation of PO-PLS with manual and automatic component selections.
popls( X, Y, commons = 2, auto = TRUE, auto.par = list(explVarLim = 40, rLim = 0.8), manual.par = list(ncomp = rep(0, length(X)), ncommon = list()) )popls( X, Y, commons = 2, auto = TRUE, auto.par = list(explVarLim = 40, rLim = 0.8), manual.par = list(ncomp = rep(0, length(X)), ncommon = list()) )
X |
|
Y |
|
commons |
|
auto |
|
auto.par |
|
manual.par |
|
PO-PLS decomposes a set of input data blocks into common, local and distinct components
through a process involving pls and gca. The rLim parameter is
a lower bound for the GCA correlation when building common components, while explVarLim is the minimum
explained variance for common components and unique components.
A multiblock object with block-wise, local and common loadings and scores. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
I Måge, BH Mevik, T Næs. (2008). Regression models with process variables and parallel blocks of raw material measurements. Journal of Chemometrics: A Journal of the Chemometrics Society 22 (8), 443-456
I Måge, E Menichelli, T Næs (2012). Preference mapping by PO-PLS: Separating common and unique information in several data blocks. Food quality and preference 24 (1), 8-16
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) # Automatic analysis pot.po.auto <- popls(potato[1:3], potato[['Sensory']][,1]) pot.po.auto$explVar # Manual choice of up to 5 components for each block and 1, 0, and 2 blocks, # respectively from the (1,2), (1,3) and (2,3) combinations of blocks. pot.po.man <- popls(potato[1:3], potato[['Sensory']][,1], auto=FALSE, manual.par = list(ncomp=c(5,5,5), ncommon=c(1,0,2))) pot.po.man$explVar # Score plot for local (2,3) components plot(scores(pot.po.man,3), comps=1:2, labels="names")data(potato) # Automatic analysis pot.po.auto <- popls(potato[1:3], potato[['Sensory']][,1]) pot.po.auto$explVar # Manual choice of up to 5 components for each block and 1, 0, and 2 blocks, # respectively from the (1,2), (1,3) and (2,3) combinations of blocks. pot.po.man <- popls(potato[1:3], potato[['Sensory']][,1], auto=FALSE, manual.par = list(ncomp=c(5,5,5), ncommon=c(1,0,2))) pot.po.man$explVar # Score plot for local (2,3) components plot(scores(pot.po.man,3), comps=1:2, labels="names")
A dataset containing 9 blocks of measurements on 26 potatoes. Original dataset can be found at http://models.life.ku.dk/Texture_Potatoes. This version has been pre-processed as follows (corresponding to Liland et al. 2016):
Variables containing NaN have been removed.
Chemical and Compression blocks have been scaled by standard deviations.
NIR blocks have been subjected to SNV (Standard Normal Variate).
data(potato)data(potato)
A data.frame having 26 rows and 9 variables:
Matrix of chemical measurements
Matrix of rheological compression data
Matrix of near-infrared measurements of raw potatoes
Matrix of near-infrared measurements of cooked potatoes
Matrix of NMR (CPMG) measurements of raw potatoes
Matrix of NMR (CPMG) measurements of cooked potatoes
Matrix of NMR (FID) measurements of raw potatoes
Matrix of NMR (FID) measurements of cooked potatoes
Matrix of sensory assessments
L.G.Thygesen, A.K.Thybo, S.B.Engelsen, Prediction of Sensory Texture Quality of Boiled Potatoes From Low-field1H NMR of Raw Potatoes. The Role of Chemical Constituents. LWT - Food Science and Technology 34(7), 2001, pp 469-477.
Kristian Hovde Liland, Tormod Næs, Ulf Geir Indahl, ROSA – a fast extension of Partial Least Squares Regression for Multiblock Data Analysis, Journal of Chemometrics 30:11 (2016), pp. 651-662.
Prediction for the mbpls (MBPLS) model. New responses or scores are predicted using a fitted model and a data.frame or list containing matrices of observations.
## S3 method for class 'mbpls' predict( object, newdata, ncomp = 1:object$ncomp, comps, type = c("response", "scores"), na.action = na.pass, ... )## S3 method for class 'mbpls' predict( object, newdata, ncomp = 1:object$ncomp, comps, type = c("response", "scores"), na.action = na.pass, ... )
object |
an |
newdata |
a data frame. The new data. If missing, the training data is used. |
ncomp, comps
|
vector of positive integers. The components to use in the prediction. See below. |
type |
character. Whether to predict scores or response values |
na.action |
function determining what should be done with missing
values in |
... |
further arguments. Currently not used |
When type is "response" (default), predicted response values
are returned. If comps is missing (or is NULL), predictions
for length(ncomp) models with ncomp[1] components,
ncomp[2] components, etc., are returned. Otherwise, predictions for
a single model with the exact components in comps are returned.
(Note that in both cases, the intercept is always included in the
predictions. It can be removed by subtracting the Ymeans component
of the fitted model.)
When type is "scores", predicted score values are returned for
the components given in comps. If comps is missing or
NULL, ncomps is used instead.
When type is "response", a three dimensional array of
predicted response values is returned. The dimensions correspond to the
observations, the response variables and the model sizes, respectively.
When type is "scores", a score matrix is returned.
A warning message like ‘'newdata' had 10 rows but variable(s)
found have 106 rows’ means that not all variables were found in the
newdata data frame. This (usually) happens if the formula contains
terms like yarn$NIR. Do not use such terms; use the data
argument instead. See mvr for details.
Kristian Hovde Liland
data(potato) mb <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, subset = 1:26 <= 18) testdata <- subset(potato, 1:26 > 18) # Predict response yhat <- predict(mb, newdata = testdata) # Predict scores and plot scores <- predict(mb, newdata = testdata, type = "scores") scoreplot(mb) points(scores[,1], scores[,2], col="red") legend("topright", legend = c("training", "test"), col=1:2, pch = 1)data(potato) mb <- mbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, subset = 1:26 <= 18) testdata <- subset(potato, 1:26 > 18) # Predict response yhat <- predict(mb, newdata = testdata) # Predict scores and plot scores <- predict(mb, newdata = testdata, type = "scores") scoreplot(mb) points(scores[,1], scores[,2], col="red") legend("topright", legend = c("training", "test"), col=1:2, pch = 1)
This is an interface to simplify preprocessing of one, a subset or all
blocks in a multiblock object, e.g., a data.frame (see block.data.frame)
or list. Several standard preprocessing methods are supplied in addition to
letting the user supply it's own function.
block.preprocess( X, block = 1:length(X), fun = c("autoscale", "center", "scale", "SNV", "EMSC", "Fro", "FroSq", "SingVal"), ... )block.preprocess( X, block = 1:length(X), fun = c("autoscale", "center", "scale", "SNV", "EMSC", "Fro", "FroSq", "SingVal"), ... )
X |
|
block |
|
fun |
|
... |
additional arguments to underlying functions. |
The fun parameter controls the type of preprocessing to be performed:
autoscale: centre and scale each feature/variable.
center: centre each feature/variable.
scale: scale each feature/variable.
SNV: Standard Normal Variate correction, i.e., centre and scale each sample across features/variables.
EMSC: Extended Multiplicative Signal Correction defaulting to basic EMSC (2nd order polynomials). Further parameters are sent to EMSC::EMSC.
Fro: Frobenius norm scaling of whole block.
FroSq: Squared Frobenius norm scaling of whole block (sum of squared values).
SingVal: Singular value scaling of whole block (first singular value).
User defined: If a function is supplied, this will be applied to chosen blocks. Preprocessing can be done for all blocks or a subset. It can also be done in a series of operations to combine preprocessing techniques.
The input multiblock object is preprocessed and returned.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) # Autoscale Chemical block potato <- block.preprocess(potato, block = "Chemical", "autoscale") # Apply SNV to NIR blocks potato <- block.preprocess(potato, block = 3:4, "SNV") # Centre Sensory block potato <- block.preprocess(potato, block = "Sensory", "center") # Scale all blocks to unit Frobenius norm potato <- block.preprocess(potato, fun = "Fro") # Effect of SNV NIR <- (potato$NIRraw + rnorm(26)) * rnorm(26,1,0.2) NIRc <- block.preprocess(list(NIR), fun = "SNV")[[1]] old.par <- par(mfrow = c(2,1), mar = c(4,4,1,1)) matplot(t(NIR), type="l", main = "uncorrected", ylab = "") matplot(t(NIRc), type="l", main = "corrected", ylab = "") par(old.par)data(potato) # Autoscale Chemical block potato <- block.preprocess(potato, block = "Chemical", "autoscale") # Apply SNV to NIR blocks potato <- block.preprocess(potato, block = 3:4, "SNV") # Centre Sensory block potato <- block.preprocess(potato, block = "Sensory", "center") # Scale all blocks to unit Frobenius norm potato <- block.preprocess(potato, fun = "Fro") # Effect of SNV NIR <- (potato$NIRraw + rnorm(26)) * rnorm(26,1,0.2) NIRc <- block.preprocess(list(NIR), fun = "SNV")[[1]] old.par <- par(mfrow = c(2,1), mar = c(4,4,1,1)) matplot(t(NIR), type="l", main = "uncorrected", ylab = "") matplot(t(NIRc), type="l", main = "corrected", ylab = "") par(old.par)
Formula based interface to the ROSA algorithm following the style of the pls package.
rosa( formula, ncomp, Y.add, common.comp = 1, data, subset, na.action, scale = FALSE, weights = NULL, validation = c("none", "CV", "LOO"), internal.validation = FALSE, fixed.block = NULL, design.block = NULL, canonical = TRUE, ... )rosa( formula, ncomp, Y.add, common.comp = 1, data, subset, na.action, scale = FALSE, weights = NULL, validation = c("none", "CV", "LOO"), internal.validation = FALSE, fixed.block = NULL, design.block = NULL, canonical = TRUE, ... )
formula |
Model formula accepting a single response (block) and predictor block names separated by + signs. |
ncomp |
The maximum number of ROSA components. |
Y.add |
Optional response(s) available in the data set. |
common.comp |
Automatically create all combinations of common components up to length |
data |
The data set to analyse. |
subset |
Expression for subsetting the data before modelling. |
na.action |
How to handle NAs (no action implemented). |
scale |
Optionally scale predictor variables by their individual standard deviations. |
weights |
Optional object weights. |
validation |
Optional cross-validation strategy "CV" or "LOO". |
internal.validation |
Optional cross-validation for block selection process, "LOO", "CV3", "CV5", "CV10" (CV-number of segments), or vector of integers (default = FALSE). |
fixed.block |
integer vector with block numbers for each component (0 = not fixed) or list of length <= ncomp (element length 0 = not fixed). |
design.block |
integer vector containing block numbers of design blocks |
canonical |
logical indicating if canonical correlation should be use when calculating loading weights (default), enabling B/W maximization, common components, etc. Alternatively (FALSE) a PLS2 strategy, e.g. for spectra response, is used. |
... |
Additional arguments for |
ROSA is an opportunistic method sequentially selecting components from whichever block explains the response most effectively. It can be formulated as a PLS model on concatenated input block with block selection per component. This implementation adds several options that are not described in the literature. Most importantly, it opens for internal validation in the block selection process, making this more robust. In addition it handles design blocks explicitly, enables classification and secondary responses (CPLS), and definition of common components.
An object of classes rosa and mvr having several associated printing (rosa_results) and plotting methods (rosa_plots).
Liland, K.H., Næs, T., and Indahl, U.G. (2016). ROSA - a fast extension of partial least squares regression for multiblock data analysis. Journal of Chemometrics, 30, 651–662, doi:10.1002/cem.2824.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in rosa_results and rosa_plots, respectively.
data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 10, validation = "CV", segments = 5) summary(mod) # For examples of ROSA results and plotting see # ?rosa_results and ?rosa_plots.data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 10, validation = "CV", segments = 5) summary(mod) # For examples of ROSA results and plotting see # ?rosa_results and ?rosa_plots.
Various plotting procedures for rosa objects.
## S3 method for class 'rosa' image( x, type = c("correlation", "residual", "order"), ncomp = x$ncomp, col = mcolors(128), legend = TRUE, mar = c(5, 6, 4, 7), las = 1, ... ) ## S3 method for class 'rosa' barplot( height, type = c("train", "CV"), ncomp = height$ncomp, col = mcolors(ncomp), ... )## S3 method for class 'rosa' image( x, type = c("correlation", "residual", "order"), ncomp = x$ncomp, col = mcolors(128), legend = TRUE, mar = c(5, 6, 4, 7), las = 1, ... ) ## S3 method for class 'rosa' barplot( height, type = c("train", "CV"), ncomp = height$ncomp, col = mcolors(ncomp), ... )
x |
A |
type |
An optional |
ncomp |
Integer to control the number of components to plot (if fewer than the fitted number of components). |
col |
Colours used for the image and bar plot, defaulting to mcolors(128). |
legend |
Logical indicating if a legend should be included (default = TRUE) for |
mar |
Figure margins, default = c(5,6,4,7) for |
las |
Axis text direction, default = 1 for |
... |
Additional parameters passed to |
height |
A |
Usage of the functions are shown using generics in the examples below. image.rosa
makes an image plot of each candidate score's correlation to the winner or the block-wise
response residual. These plots can be used to find alternative block selection for tweaking
the ROSA model. barplot.rosa makes barplot of block and component explained variances.
loadingweightsplot is an adaptation of pls::loadingplot to plot loading weights.
No return.
Liland, K.H., Næs, T., and Indahl, U.G. (2016). ROSA - a fast extension of partial least squares regression for multiblock data analysis. Journal of Chemometrics, 30, 651–662, doi:10.1002/cem.2824.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results in rosa_results.
data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 5) image(mod) barplot(mod) loadingweightplot(mod)data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 5) image(mod) barplot(mod) loadingweightplot(mod)
Standard result computation and extraction functions for ROSA (rosa).
## S3 method for class 'rosa' predict( object, newdata, ncomp = 1:object$ncomp, comps, type = c("response", "scores"), na.action = na.pass, ... ) ## S3 method for class 'rosa' coef(object, ncomp = object$ncomp, comps, intercept = FALSE, ...) ## S3 method for class 'rosa' print(x, ...) ## S3 method for class 'rosa' summary( object, what = c("all", "validation", "training"), digits = 4, print.gap = 2, ... ) blockexpl(object, ncomp = object$ncomp, type = c("train", "CV")) ## S3 method for class 'rosaexpl' print(x, digits = 3, compwise = FALSE, ...) rosa.classify(object, classes, newdata, ncomp, LQ) ## S3 method for class 'rosa' scores(object, ...) ## S3 method for class 'rosa' loadings(object, ...)## S3 method for class 'rosa' predict( object, newdata, ncomp = 1:object$ncomp, comps, type = c("response", "scores"), na.action = na.pass, ... ) ## S3 method for class 'rosa' coef(object, ncomp = object$ncomp, comps, intercept = FALSE, ...) ## S3 method for class 'rosa' print(x, ...) ## S3 method for class 'rosa' summary( object, what = c("all", "validation", "training"), digits = 4, print.gap = 2, ... ) blockexpl(object, ncomp = object$ncomp, type = c("train", "CV")) ## S3 method for class 'rosaexpl' print(x, digits = 3, compwise = FALSE, ...) rosa.classify(object, classes, newdata, ncomp, LQ) ## S3 method for class 'rosa' scores(object, ...) ## S3 method for class 'rosa' loadings(object, ...)
object |
A |
newdata |
Optional new data with the same types of predictor blocks as the ones used for fitting the object. |
ncomp |
An |
comps |
An |
type |
For |
na.action |
Function determining what to do with missing values in |
... |
Additional arguments. Currently not implemented. |
intercept |
A |
x |
A |
what |
A |
digits |
The number of digits used for printing. |
print.gap |
Gap between columns when printing. |
compwise |
Logical indicating if block-wise (default/FALSE) or component-wise (TRUE) explained variance should be printed. |
classes |
A |
LQ |
A |
Usage of the functions are shown using generics in the examples below.
Prediction, regression coefficients, object printing and summary are available through:
predict.rosa, coef.rosa, print.rosa and summary.rosa.
Explained variances are available (block-wise and global) through blockexpl and print.rosaexpl.
Scores and loadings have their own extensions of scores() and loadings() throught
scores.rosa and loadings.rosa. Finally, there is work in progress on classifcation
support through rosa.classify.
When type is "response" (default), predicted response values
are returned. If comps is missing (or is NULL), predictions
for length(ncomp) models with ncomp[1] components,
ncomp[2] components, etc., are returned. Otherwise, predictions for
a single model with the exact components in comps are returned.
(Note that in both cases, the intercept is always included in the
predictions. It can be removed by subtracting the Ymeans component
of the fitted model.)
If comps is missing (or is NULL), coef()[,,ncomp[i]]
are the coefficients for models with ncomp[i] components, for . Also, if intercept = TRUE, the first
dimension is , with the intercept coefficients as the first
row.
If comps is given, however, coef()[,,comps[i]] are the
coefficients for a model with only the component comps[i], i.e., the
contribution of the component comps[i] on the regression
coefficients.
Returns depend on method used, e.g. predict.rosa returns predicted responses
or scores depending on inputs, coef.rosa returns regression coefficients, blockexpl
returns an object of class rosaexpl containing block-wise and component-wise explained variance contained in a matrix with attributes.
Liland, K.H., Næs, T., and Indahl, U.G. (2016). ROSA - a fast extension of partial least squares regression for multiblock data analysis. Journal of Chemometrics, 30, 651–662, doi:10.1002/cem.2824.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in rosa_results and rosa_plots, respectively.
data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 5, subset = 1:20) testset <- potato[-(1:20),]; testset$Sensory <- testset$Sensory[,1,drop=FALSE] predict(mod, testset, ncomp=5) dim(coef(mod, ncomp=5)) # <variables x responses x components> print(mod) summary(mod) blockexpl(mod) print(blockexpl(mod), compwise=TRUE)data(potato) mod <- rosa(Sensory[,1] ~ ., data = potato, ncomp = 5, subset = 1:20) testset <- potato[-(1:20),]; testset$Sensory <- testset$Sensory[,1,drop=FALSE] predict(mod, testset, ncomp=5) dim(coef(mod, ncomp=5)) # <variables x responses x components> print(mod) summary(mod) blockexpl(mod) print(blockexpl(mod), compwise=TRUE)
This is a basic implementation of the SCA-P algorithm (least restricted SCA) with support for both sample- and variable-linked modes.
sca(X, ncomp = 2, scale = FALSE, samplelinked = "auto", ...)sca(X, ncomp = 2, scale = FALSE, samplelinked = "auto", ...)
X |
|
ncomp |
|
scale |
|
samplelinked |
|
... |
additional arguments (not used). |
SCA, in its original variable-linked version, calculates common loadings and block-wise scores. There are many possible constraints and specialisations. This implementations uses PCA as the backbone, thus resulting in deterministic, ordered components. A parameter controls the linking mode, but if left untouched an attempt is made at automatically determining variable or sample linking.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Levin, J. (1966) Simultaneous factor analysis of several gramian matrices. Psychometrika, 31(3), 413–419.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
# Object linked data data(potato) potList <- as.list(potato[c(1,2,9)]) pot.sca <- sca(potList) plot(scores(pot.sca), labels="names") # Variable linked data data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) pot.sca <- sca(candyList, samplelinked = FALSE) pot.sca# Object linked data data(potato) potList <- as.list(potato[c(1,2,9)]) pot.sca <- sca(potList) plot(scores(pot.sca), labels="names") # Variable linked data data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) pot.sca <- sca(candyList, samplelinked = FALSE) pot.sca
A dataset containing simulated data for 4 connected events where A is the
starting point and D is the end point. This can be described as a directed
acyclic graph (sketched below, moving left->right).
Subpaths include: ABD, AD, ABCD, ACD
data(simulated)data(simulated)
A list of matrices having 200 rows and 10 variables:
Simulated matrix A
Simulated matrix B
...
Tormod Næs, Rosaria Romano, Oliver Tomic, Ingrid Måge, Age Smilde, Kristian Hovde Liland, Sequential and orthogonalized PLS (SO-PLS) regression for path analysis: Order of blocks and relations between effects. Journal of Chemometrics, In Press
sMB-PLS is an adaptation of MB-PLS (mbpls) that enforces sparseness in loading weights
when computing PLS components in the global model.
smbpls( formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, scale = FALSE, shrink = NULL, truncation = NULL, trunc.width = 0.95, blockScale = c("sqrtnvar", "ssq", "none"), ... )smbpls( formula, data, subset, na.action, X = NULL, Y = NULL, ncomp = 1, scale = FALSE, shrink = NULL, truncation = NULL, trunc.width = 0.95, blockScale = c("sqrtnvar", "ssq", "none"), ... )
formula |
Model formula accepting a single response (block) and predictor block names separated by + signs. |
data |
The data set to analyse. |
subset |
Expression for subsetting the data before modelling. |
na.action |
How to handle NAs (no action implemented). |
X |
|
Y |
|
ncomp |
|
scale |
|
shrink |
|
truncation |
|
trunc.width |
|
blockScale |
Either a |
... |
additional arguments to pls::plsr. |
Two versions of sparseness are supplied: Soft-Threshold PLS, also known as Sparse PLS, and Truncation PLS. The former uses L1 shrinkage of loading weights, while the latter comes in two flavours, both estimating inliers and outliers. The "Lenth" method uses asymmetric confidence intervals around the median of a loading weigh vector to estimate inliers. The "quantile" method uses a quantile plot approach to estimate outliers as deviations from the estimated quantile line. As with ordinary MB-PLS scaled input blocks (1/sqrt(ncol)) are used.
Block weighting is performed after scaling all variables and is by default
"sqrtnvar": 1/sqrt(ncol(X[[i]])) in each block. Alternatives
are "ssq": 1/norm(X[[i]], "F")^2 and "none": 1/1. Finally, if
a numeric vector is supplied, it will be used to scale the blocks
after "ssq" scaling, i.e., Z[[i]] = X[[i]] / norm(X[[i]], "F")^2 * blockScale[i].
multiblock, mvr object with super-scores, super-loadings, block-scores and block-loading, and the underlying
mvr (PLS) object for the super model, with all its result and plot possibilities. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Sæbø, S.; Almøy, T.; Aarøe, J. & Aastveit, A. ST-PLS: a multi-directional nearest shrunken centroid type classifier via PLS Journal of Chemometrics: A Journal of the Chemometrics Society, Wiley Online Library, 2008, 22, 54-62.
Lê Cao, K.; Rossouw, D.; Robert-Granié, C. & Besse, P. A sparse PLS for variable selection when integrating omics data Statistical applications in genetics and molecular biology, 2008, 7.
Liland, K.; Høy, M.; Martens, H. & Sæbø, S. Distribution based truncation for variable selection in subspace methods for multivariate regression Chemometrics and Intelligent Laboratory Systems, 2013, 122, 103-111.
Karaman, I.; Nørskov, N.; Yde, C.; Hedemann, M.; Knudsen, K. & Kohler, A. Sparse multi-block PLSR for biomarker discovery when integrating data from LC–MS and NMR metabolomics Metabolomics, 2015, 11, 367-379.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(potato) # Truncation MB-PLS # Loading weights inside 60% confidence intervals around the median are set to 0. tmb <- smbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, truncation = "Lenth", trunc.width = 0.6) # Alternative XY-interface tmb.XY <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, truncation = "Lenth", trunc.width = 0.6) identical(tmb, tmb.XY) scoreplot(tmb, labels="names") # Exploiting mvr object structure from pls package loadingweightplot(tmb, labels="names") # Soft-Threshold / Sparse MB-PLS # Loading weights are subtracted by 60% of maximum value. smb <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, shrink = 0.6) print(smb) scoreplot(smb, labels="names") # Exploiting mvr object structure from pls package loadingweightplot(smb, labels="names") # Emphasis may be different for blocks smb <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, shrink = 0.6, blockScale = c(1, 10))data(potato) # Truncation MB-PLS # Loading weights inside 60% confidence intervals around the median are set to 0. tmb <- smbpls(Sensory ~ Chemical+Compression, data=potato, ncomp = 5, truncation = "Lenth", trunc.width = 0.6) # Alternative XY-interface tmb.XY <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, truncation = "Lenth", trunc.width = 0.6) identical(tmb, tmb.XY) scoreplot(tmb, labels="names") # Exploiting mvr object structure from pls package loadingweightplot(tmb, labels="names") # Soft-Threshold / Sparse MB-PLS # Loading weights are subtracted by 60% of maximum value. smb <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, shrink = 0.6) print(smb) scoreplot(smb, labels="names") # Exploiting mvr object structure from pls package loadingweightplot(smb, labels="names") # Emphasis may be different for blocks smb <- smbpls(X=potato[c('Chemical','Compression')], Y=potato[['Sensory']], ncomp = 5, shrink = 0.6, blockScale = c(1, 10))
SO-PLS-PM is the use of SO-PLS for path-modelling. This particular function is used to compute effects (explained variances) in sub-paths of the directed acyclic graph.
sopls_pm( X, Y, ncomp, max_comps = min(sum(ncomp), 20), sel.comp = "opt", computeAdditional = FALSE, sequential = FALSE, B = NULL, k = 10, type = "consecutive", simultaneous = TRUE ) ## S3 method for class 'SO_TDI' print(x, showComp = TRUE, heading = "SO-PLS path effects", digits = 2, ...) sopls_pm_multiple( X, ncomp, max_comps = min(sum(ncomp), 20), sel.comp = "opt", computeAdditional = FALSE, sequential = FALSE, B = NULL, k = 10, type = "consecutive" ) ## S3 method for class 'SO_TDI_multiple' print(x, heading = "SO-PLS path effects", digits = 2, ...)sopls_pm( X, Y, ncomp, max_comps = min(sum(ncomp), 20), sel.comp = "opt", computeAdditional = FALSE, sequential = FALSE, B = NULL, k = 10, type = "consecutive", simultaneous = TRUE ) ## S3 method for class 'SO_TDI' print(x, showComp = TRUE, heading = "SO-PLS path effects", digits = 2, ...) sopls_pm_multiple( X, ncomp, max_comps = min(sum(ncomp), 20), sel.comp = "opt", computeAdditional = FALSE, sequential = FALSE, B = NULL, k = 10, type = "consecutive" ) ## S3 method for class 'SO_TDI_multiple' print(x, heading = "SO-PLS path effects", digits = 2, ...)
X |
A |
Y |
A |
ncomp |
An |
max_comps |
Maximum total number of components. |
sel.comp |
A |
computeAdditional |
A |
sequential |
A |
B |
An |
k |
An |
type |
A |
simultaneous |
|
x |
An object of type |
showComp |
A |
heading |
A |
digits |
An |
... |
Not implemented |
sopls_pm computes 'total', 'direct', 'indirect' and 'additional' effects for the 'first' versus the
'last' input block by cross-validated explained variances. 'total' is the explained variance when doing
regression of 'first' -> 'last'. 'indirect' is the the same, but controlled for the intermediate blocks.
'direct' is the left-over part of the 'total' explained variance when subtracting the 'indirect'. Finally,
'additional' is the added explained variance of 'last' for each block following 'first'.
sopls_pm_multiple is a wrapper for sopls_pm that repeats the calculation for all pairs of blocks
from 'first' to 'last'. Where sopls_pm has a separate response, Y, signifying the 'last' block,
sopls_pm_multiple has multiple 'last' blocks, depending on sub-path, thus collects the response(s)
from the list of blocks X.
When sel.comp = "opt", the number of components for all models are optimized using cross-validation within the ncomp and max_comps supplied. If sel.comp is "chi", an optimization is also performed, but parsimonious solutions are sought through a chi-square chriterion. When setting sel.comp to a numeric vector, exact selection of number of components is performed.
When setting B to a number, e.g. 200, the procedures above are repeated B times using bootstrapping to estimate standard deviations of the cross-validated explained variances.
An object of type SO_TDI containing total, direct and indirect effects, plus
possibly additional effects and standard deviations (estimated by bootstrapping).
Menichelli, E., Almøy, T., Tomic, O., Olsen, N. V., & Næs, T. (2014). SO-PLS as an exploratory tool for path modelling. Food quality and preference, 36, 122-134.
Næs, T., Romano, R., Tomic, O., Måge, I., Smilde, A., & Liland, K. H. (2020). Sequential and orthogonalized PLS (SO-PLS) regression for path analysis: Order of blocks and relations between effects. Journal of Chemometrics, e3243.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
# Single path for the potato data: data(potato) pot.pm <- sopls_pm(potato[1:3], potato[['Sensory']], c(5,5,5), computeAdditional=TRUE) pot.pm # Corresponding SO-PLS model: # so <- sopls(Sensory ~ ., data=potato[c(1,2,3,9)], ncomp=c(5,5,5), validation="CV", segments=10) # maageSeq(pot.so, compSeq = c(3,2,4)) # All path in the forward direction for the wine data: data(wine) pot.pm.multiple <- sopls_pm_multiple(wine, ncomp = c(4,2,9,8)) pot.pm.multiple# Single path for the potato data: data(potato) pot.pm <- sopls_pm(potato[1:3], potato[['Sensory']], c(5,5,5), computeAdditional=TRUE) pot.pm # Corresponding SO-PLS model: # so <- sopls(Sensory ~ ., data=potato[c(1,2,3,9)], ncomp=c(5,5,5), validation="CV", segments=10) # maageSeq(pot.so, compSeq = c(3,2,4)) # All path in the forward direction for the wine data: data(wine) pot.pm.multiple <- sopls_pm_multiple(wine, ncomp = c(4,2,9,8)) pot.pm.multiple
Function for computing standard SO-PLS based on the interface of the pls package.
sopls( formula, ncomp, max_comps = min(sum(ncomp), 20), data, subset, na.action, scale = FALSE, validation = c("none", "CV", "LOO"), sequential = FALSE, segments = 10, sel.comp = "opt", progress = TRUE, ... )sopls( formula, ncomp, max_comps = min(sum(ncomp), 20), data, subset, na.action, scale = FALSE, validation = c("none", "CV", "LOO"), sequential = FALSE, segments = 10, sel.comp = "opt", progress = TRUE, ... )
formula |
Model formula accepting a single response (block) and predictor block names separated by + signs. |
ncomp |
Numeric vector of components per block or scalar of overall maximum components. |
max_comps |
Maximum total number of components from all blocks combined (<= sum(ncomp)). |
data |
The data set to analyse. |
subset |
Expression for subsetting the data before modelling. |
na.action |
How to handle NAs (no action implemented). |
scale |
Logical indicating if variables should be scaled. |
validation |
Optional cross-validation strategy "CV" or "LOO". |
sequential |
Logical indicating if optimal components are chosen sequentially or globally (default=FALSE). |
segments |
Optional number of segments or list of segments for cross-validation. (See |
sel.comp |
Character indicating if sequential optimal number of components should be chosen as minimum RMSECV ('opt', default) or by Chi-square test ('chi'). |
progress |
Logical indicating if a progress bar should be displayed while cross-validating. |
... |
Additional arguments to underlying methods. |
SO-PLS is a method which handles two or more input blocks by sequentially performing PLS on blocks against a response and orthogonalising the remaining blocks on the extracted components. Component number optimisation can either be done globally (best combination across blocks) or sequentially (determine for one block, move to next, etc.).
An sopls, mvr object with scores, loadings, etc.
associated with printing (sopls_results) and plotting methods (sopls_plots).
Jørgensen K, Mevik BH, Næs T. Combining designed experiments with several blocks of spectroscopic data. Chemometr Intell Lab Syst. 2007;88(2): 154–166.
SO-PLS result functions, sopls_results, SO-PLS plotting functions, sopls_plots, SO-PLS Måge plot, maage, and SO-PLS path-modelling, SO_TDI.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
data(potato) so <- sopls(Sensory ~ Chemical + Compression, data=potato, ncomp=c(10,10), max_comps=10, validation="CV", segments=10) summary(so) # Scatter plot matrix with two first components from Chemical block # and 1 component from the Compression block. scoreplot(so, comps=list(1:2,1), ncomp=2, block=2) # Result functions and more plots for SO-PLS # are found in ?sopls_results and ?sopls_plots.data(potato) so <- sopls(Sensory ~ Chemical + Compression, data=potato, ncomp=c(10,10), max_comps=10, validation="CV", segments=10) summary(so) # Scatter plot matrix with two first components from Chemical block # and 1 component from the Compression block. scoreplot(so, comps=list(1:2,1), ncomp=2, block=2) # Result functions and more plots for SO-PLS # are found in ?sopls_results and ?sopls_plots.
Extraction of scores and loadings and adaptation of scoreplot,
loadingplot and biplot from package pls for sopls objects.
## S3 method for class 'sopls' loadings(object, ncomp = "all", block = 1, y = FALSE, ...) ## S3 method for class 'sopls' scores(object, ncomp = "all", block = 1, y = FALSE, ...) ## S3 method for class 'sopls' scoreplot( object, comps = 1:2, ncomp = NULL, block = 1, labels, identify = FALSE, type = "p", xlab, ylab, ... ) ## S3 method for class 'sopls' loadingplot( object, comps = 1:2, ncomp = NULL, block = 1, scatter = TRUE, labels, identify = FALSE, type, lty, lwd = NULL, pch, cex = NULL, col, legendpos, xlab, ylab, pretty.xlabels = TRUE, xlim, ... ) ## S3 method for class 'sopls' corrplot( object, comps = 1:2, ncomp = NULL, block = 1, labels = TRUE, col = 1:5, plotx = TRUE, ploty = TRUE, ... ) ## S3 method for class 'sopls' biplot( x, comps = 1:2, ncomp = "all", block = 1, which = c("x", "y", "scores", "loadings"), var.axes = FALSE, xlabs, ylabs, main, ... )## S3 method for class 'sopls' loadings(object, ncomp = "all", block = 1, y = FALSE, ...) ## S3 method for class 'sopls' scores(object, ncomp = "all", block = 1, y = FALSE, ...) ## S3 method for class 'sopls' scoreplot( object, comps = 1:2, ncomp = NULL, block = 1, labels, identify = FALSE, type = "p", xlab, ylab, ... ) ## S3 method for class 'sopls' loadingplot( object, comps = 1:2, ncomp = NULL, block = 1, scatter = TRUE, labels, identify = FALSE, type, lty, lwd = NULL, pch, cex = NULL, col, legendpos, xlab, ylab, pretty.xlabels = TRUE, xlim, ... ) ## S3 method for class 'sopls' corrplot( object, comps = 1:2, ncomp = NULL, block = 1, labels = TRUE, col = 1:5, plotx = TRUE, ploty = TRUE, ... ) ## S3 method for class 'sopls' biplot( x, comps = 1:2, ncomp = "all", block = 1, which = c("x", "y", "scores", "loadings"), var.axes = FALSE, xlabs, ylabs, main, ... )
object |
|
ncomp |
|
block |
|
y |
|
... |
further arguments sent to the underlying plot function(s) |
comps |
|
labels |
|
identify |
|
type |
|
xlab |
|
ylab |
|
scatter |
|
lty |
Vector of line type specifications (see |
lwd |
|
pch |
Vector of point specifications (see |
cex |
|
col |
|
legendpos |
|
pretty.xlabels |
|
xlim |
|
plotx |
|
ploty |
|
x |
|
which |
|
var.axes |
|
xlabs |
|
ylabs |
|
main |
|
If comps is supplied as a list for scoreplot, it is assumed that its elements refer to each of the
blocks up to block number block. For instance comps = list(1, 0, 1:2) will select 1 component from the first
block, no components from the second block and the first two components from the last block. This must be
matched by ncomp, specifying how many components were selected before block number block.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results are found in sopls_results.
#' @return The score and loading functions return scores and loadings, while plot functions have no return (except use of 'identify').
data(potato) so <- sopls(Sensory ~ Chemical + Compression + NIRraw, data=potato, ncomp=c(5,5,5)) # Loadings loadings(so, ncomp=c(3), block=2)[, 1:3] # Scores scores(so, block=1)[, 1:4] # Default plot from first block scoreplot(so) # Second block with names scoreplot(so, ncomp=c(3), block=2, labels="names") # Scatterplot matrix scoreplot(so, ncomp=c(3,2), block=3, comps=1:3) # Combination of blocks (see Details) scoreplot(so, ncomp=c(3,2), block=3, comps=list(1,0,1)) # Default plot from first block loadingplot(so, scatter=TRUE) # Second block with names loadingplot(so, ncomp=c(3), block=2, labels="names", scatter=TRUE) # Scatterplot matrix loadingplot(so, ncomp=c(3,2), block=3, comps=1:3, scatter=TRUE) # Correlation loadings corrplot(so, block=2, ncomp=1) # Default plot from first block biplot(so)data(potato) so <- sopls(Sensory ~ Chemical + Compression + NIRraw, data=potato, ncomp=c(5,5,5)) # Loadings loadings(so, ncomp=c(3), block=2)[, 1:3] # Scores scores(so, block=1)[, 1:4] # Default plot from first block scoreplot(so) # Second block with names scoreplot(so, ncomp=c(3), block=2, labels="names") # Scatterplot matrix scoreplot(so, ncomp=c(3,2), block=3, comps=1:3) # Combination of blocks (see Details) scoreplot(so, ncomp=c(3,2), block=3, comps=list(1,0,1)) # Default plot from first block loadingplot(so, scatter=TRUE) # Second block with names loadingplot(so, ncomp=c(3), block=2, labels="names", scatter=TRUE) # Scatterplot matrix loadingplot(so, ncomp=c(3,2), block=3, comps=1:3, scatter=TRUE) # Correlation loadings corrplot(so, block=2, ncomp=1) # Default plot from first block biplot(so)
Standard result functions for SO-PLS (sopls).
## S3 method for class 'sopls' predict( object, newdata, ncomp = object$ncomp, type = c("response", "scores"), na.action = na.pass, ... ) ## S3 method for class 'sopls' coef(object, ncomp = object$ncomp, intercept = FALSE, ...) ## S3 method for class 'sopls' print(x, ...) ## S3 method for class 'sopls' summary( object, what = c("all", "validation", "training"), digits = 4, print.gap = 2, ... ) classify(object, ...) ## S3 method for class 'sopls' classify(object, classes, newdata, ncomp, LQ = "LDA", ...) ## S3 method for class 'sopls' R2(object, estimate, newdata, ncomp = "all", individual = FALSE, ...) ## S3 method for class 'sopls' RMSEP(object, estimate, newdata, ncomp = "all", individual = FALSE, ...) pcp(object, ...) ## S3 method for class 'sopls' pcp(object, ncomp, ...) ## Default S3 method: pcp(object, X, ...) cvanova(pred, ...) ## Default S3 method: cvanova(pred, true, absRes = TRUE, ...) ## S3 method for class 'sopls' cvanova(pred, comps, absRes = TRUE, ...) ## S3 method for class 'cvanova' print(x, ...) ## S3 method for class 'cvanova' summary(object, ...) ## S3 method for class 'cvanova' plot(x, ...) ## S3 method for class 'sopls' residuals(object, ...)## S3 method for class 'sopls' predict( object, newdata, ncomp = object$ncomp, type = c("response", "scores"), na.action = na.pass, ... ) ## S3 method for class 'sopls' coef(object, ncomp = object$ncomp, intercept = FALSE, ...) ## S3 method for class 'sopls' print(x, ...) ## S3 method for class 'sopls' summary( object, what = c("all", "validation", "training"), digits = 4, print.gap = 2, ... ) classify(object, ...) ## S3 method for class 'sopls' classify(object, classes, newdata, ncomp, LQ = "LDA", ...) ## S3 method for class 'sopls' R2(object, estimate, newdata, ncomp = "all", individual = FALSE, ...) ## S3 method for class 'sopls' RMSEP(object, estimate, newdata, ncomp = "all", individual = FALSE, ...) pcp(object, ...) ## S3 method for class 'sopls' pcp(object, ncomp, ...) ## Default S3 method: pcp(object, X, ...) cvanova(pred, ...) ## Default S3 method: cvanova(pred, true, absRes = TRUE, ...) ## S3 method for class 'sopls' cvanova(pred, comps, absRes = TRUE, ...) ## S3 method for class 'cvanova' print(x, ...) ## S3 method for class 'cvanova' summary(object, ...) ## S3 method for class 'cvanova' plot(x, ...) ## S3 method for class 'sopls' residuals(object, ...)
object |
A |
newdata |
Optional new data with the same types of predictor blocks as the ones used for fitting the object. |
ncomp |
An |
type |
A |
na.action |
Function determining what to do with missing values in |
... |
Additional arguments. Currently not implemented. |
intercept |
A |
x |
A |
what |
A |
digits |
The number of digits used for printing. |
print.gap |
Gap between columns when printing. |
classes |
A |
LQ |
A |
estimate |
A |
individual |
A |
X |
A |
pred |
An object holding the CV-predicted values ( |
true |
A |
absRes |
A |
comps |
An |
The parameter ncomp controls
which components to apply/extract, resulting in the sequence of components leading up to the specific choice, i.e.
ncomp = c(2,2,1) results in the sequence 1,0,0; 2,0,0; 2,1,0; 2,2,0; 2,2,1.
Usage of the functions are shown using generics in the examples below.
Prediction, regression coefficients, object printing and summary are available through:
predict.sopls, coef.sopls, print.sopls and summary.sopls.
Explained variances and RMSEP are available through R2.sopls and RMSEP.sopls.
Principal components of predictions are available through pcp.sopls. Finally, there is work in progress on classifcation
support through classify.sopls.
Returns depend on method used, e.g. predict.sopls returns predicted responses
or scores depending on inputs, coef.sopls return regression coefficients, while print and summary methods return the object invisibly.
Jørgensen K, Mevik BH, Næs T. Combining designed experiments with several blocks of spectroscopic data. Chemometr Intell Lab Syst. 2007;88(2): 154–166.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for plotting are found in sopls_plots.
data(potato) mod <- sopls(Sensory[,1] ~ ., data = potato[c(1:3,9)], ncomp = 5, subset = 1:20) testset <- potato[-(1:20),]; testset$Sensory <- testset$Sensory[,1,drop=FALSE] predict(mod, testset, ncomp=c(2,1,2)) dim(coef(mod, ncomp=c(3,0,1))) # <variables x responses x components> R2(mod, ncomp = c(4,1,2)) print(mod) summary(mod) # PCP from sopls object modMulti <- sopls(Sensory ~ ., data = potato[c(1:3,9)], ncomp = 5, validation = "CV", segment = 5) (PCP <- pcp(modMulti, c(2,1,2))) scoreplot(PCP) # PCP from matrices preds <- modMulti$validation$Ypred[,,"2,1,2"] PCP_default <- pcp(preds, potato[1:3]) # CVANOVA modCV <- sopls(Sensory[,1] ~ ., data = potato[c(1:3,9)], ncomp = 5, validation = "CV", segment = 5) summary(cva <- cvanova(modCV, "2,1,2")) plot(cva)data(potato) mod <- sopls(Sensory[,1] ~ ., data = potato[c(1:3,9)], ncomp = 5, subset = 1:20) testset <- potato[-(1:20),]; testset$Sensory <- testset$Sensory[,1,drop=FALSE] predict(mod, testset, ncomp=c(2,1,2)) dim(coef(mod, ncomp=c(3,0,1))) # <variables x responses x components> R2(mod, ncomp = c(4,1,2)) print(mod) summary(mod) # PCP from sopls object modMulti <- sopls(Sensory ~ ., data = potato[c(1:3,9)], ncomp = 5, validation = "CV", segment = 5) (PCP <- pcp(modMulti, c(2,1,2))) scoreplot(PCP) # PCP from matrices preds <- modMulti$validation$Ypred[,,"2,1,2"] PCP_default <- pcp(preds, potato[1:3]) # CVANOVA modCV <- sopls(Sensory[,1] ~ ., data = potato[c(1:3,9)], ncomp = 5, validation = "CV", segment = 5) summary(cva <- cvanova(modCV, "2,1,2")) plot(cva)
This is a wrapper for the ade4::statis function for computing STATIS.
statis(X, ncomp = 3, scannf = FALSE, tol = 1e-07, ...)statis(X, ncomp = 3, scannf = FALSE, tol = 1e-07, ...)
X |
|
ncomp |
|
scannf |
|
tol |
|
... |
additional arguments (not used). |
STATIS is a method, related to MFA, for analysing two or more blocks. It also decomposes the data into a low-dimensional subspace but uses a different scaling of the individual blocks.
multiblock object including relevant scores and loadings. Relevant plotting functions: multiblock_plots
and result functions: multiblock_results.
Lavit, C.; Escoufier, Y.; Sabatier, R.; Traissac, P. (1994). The ACT (STATIS method). Computational Statistics & Data Analysis. 18: 97
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.statis <- statis(candyList) plot(scores(can.statis), labels="names")data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.statis <- statis(candyList) plot(scores(can.statis), labels="names")
Collection of supervised multiblock methods:
MB-PLS - Multiblock Partial Least Squares (mbpls)
sMB-PLS - Sparse Multiblock Partial Least Squares (smbpls)
SO-PLS - Sequential and Orthogonalized PLS (sopls)
PO-PLS - Parallel and Orthogonalized PLS (popls)
ROSA - Response Oriented Sequential Alternation (rosa)
mbRDA - Multiblock Redundancy Analysis (mbrda)
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
data(potato) mb <- mbpls(Sensory ~ Chemical + Compression, data=potato, ncomp = 5) print(mb) # Convert data.frame with AsIs objects to list of matrices potatoList <- lapply(potato, unclass) mbr <- mbrda(Sensory ~ Chemical + Compression, data=potatoList, ncomp = 10) print(mbr) scoreplot(mbr, labels="names")data(potato) mb <- mbpls(Sensory ~ Chemical + Compression, data=potato, ncomp = 5) print(mb) # Convert data.frame with AsIs objects to list of matrices potatoList <- lapply(potato, unclass) mbr <- mbrda(Sensory ~ Chemical + Compression, data=potatoList, ncomp = 10) print(mbr) scoreplot(mbr, labels="names")
Compute a list of all possible block combinations where
the number of blocks in each combination is limited by parameters
min_level and max_level.
unique_combos(n_block, max_level, min_level = 2)unique_combos(n_block, max_level, min_level = 2)
n_block |
|
max_level |
|
min_level |
|
This function is used for minimal redundancy implementations of
rosa and sopls and for lookups into computed
components.
A list of unique block combinations.
unique_combos(3, 2)unique_combos(3, 2)
Collection of unsupervised multiblock methods:
SCA - Simultaneous Component Analysis (sca)
GCA - Generalized Canonical Analysis (gca)
GPA - Generalized Procrustes Analysis (gpa)
MFA - Multiple Factor Analysis (mfa)
PCA-GCA (pcagca)
DISCO - Distinctive and Common Components with SCA (disco)
HPCA - Hierarchical Principal component analysis (hpca)
MCOA - Multiple Co-Inertia Analysis (mcoa)
JIVE - Joint and Individual Variation Explained (jive)
STATIS - Structuration des Tableaux à Trois Indices de la Statistique (statis)
HOGSVD - Higher Order Generalized SVD (hogsvd)
Original documentation of STATIS: statis. JIVE, STATIS and HOGSVD assume variable linked matrices/data.frames, while SCA handles both links.
Overviews of available methods, multiblock, and methods organised by main structure: basic, unsupervised, asca, supervised and complex.
Common functions for computation and extraction of results and plotting are found in multiblock_results and multiblock_plots, respectively.
# Object linked data data(potato) potList <- as.list(potato[c(1,2,9)]) pot.sca <- sca(potList) # Variable linked data data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.statis <- statis(candyList) plot(can.statis$statis)# Object linked data data(potato) potList <- as.list(potato[c(1,2,9)]) pot.sca <- sca(potList) # Variable linked data data(candies) candyList <- lapply(1:nlevels(candies$candy),function(x)candies$assessment[candies$candy==x,]) can.statis <- statis(candyList) plot(can.statis$statis)
This dataset contains sensory assessment of 21 wines. The assessments are grouped according to the tasting process and thus have a natural ordering with a all blocks pointing forward to all remaining blocks in the process.
data(wine)data(wine)
A data.frame having 21 rows and 5 variables:
Matrix of sensory assessments
Matrix of sensory assessments
Matrix of sensory assessments
Matrix of sensory assessments
Matrix of sensory assessments
Escofier B, Pages L. Analyses Factorielles Simples and Multiples. Paris: Dunod; 1988.