wrapper.sgcca {mixOmics} | R Documentation |
Wrapper function to perform Sparse Generalised Canonical Correlation Analysis (sGCCA), a generalised approach for the integration of multiple datasets. For more details, see the help(sgcca)
from the RGCCA package.
wrapper.sgcca(X, design = 1 - diag(length(X)), penalty = NULL, ncomp = 1, keepX, scheme = "horst", mode="canonical", scale = TRUE, init = "svd.single", tol = .Machine$double.eps, max.iter=1000, near.zero.var = FALSE, all.outputs = TRUE)
X |
a list of data sets (called 'blocks') matching on the same samples. Data in the list should be arranged in samples x variables. |
design |
numeric matrix of size (number of blocks in X) x (number of blocks in X) with values between 0 and 1. Each value indicates the strenght of the relationship to be modelled between two blocks using sGCCA; a value of 0 indicates no relationship, 1 is the maximum value. If |
penalty |
numeric vector of length the number of blocks in |
ncomp |
the number of components to include in the model. Default to 1. |
keepX |
A vector of same length as X. Each entry keepX[i] is the number of X[[i]]-variables kept in the model. |
scheme |
Either "horst", "factorial" or "centroid" (Default: "horst"). |
mode |
character string. What type of algorithm to use, (partially) matching
one of |
scale |
boleean. If scale = TRUE, each block is standardized to zero means and unit variances (default: TRUE) |
init |
Mode of initialization use in the algorithm, either by Singular Value Decompostion of the product of each block of X with Y ("svd") or each block independently ("svd.single") . Default to "svd.single". |
tol |
Convergence stopping value. |
max.iter |
integer, the maximum number of iterations. |
near.zero.var |
boolean, see the internal |
all.outputs |
boolean. Computation can be faster when some specific (and non-essential) outputs are not calculated. Default = |
This wrapper function performs sGCCA (see RGCCA) with 1, … ,ncomp
components on each block data set.
A supervised or unsupervised model can be run. For a supervised model, the unmap
function should be used as an input data set.
More details can be found on the package RGCCA.
Note that this function is the same as block.spls
with different default arguments.
More details about the PLS modes in ?pls
.
wrapper.sgcca
returns an object of class "sgcca"
, a list
that contains the following components:
data |
the input data set (as a list). |
design |
the input design. |
variates |
the sgcca components. |
loadings |
the loadings for each block data set (outer wieght vector). |
loadings.star |
the laodings, standardised. |
penalty |
the input penalty parameter. |
scheme |
the input schme. |
ncomp |
the number of components included in the model for each block. |
crit |
the convergence criterion. |
AVE |
Indicators of model quality based on the Average Variance Explained (AVE): AVE(for one block), AVE(outer model), AVE(inner model).. |
names |
list containing the names to be used for individuals and variables. |
More details can be found in the references.
Arthur Tenenhaus, Vincent Guillemot and Kim-Anh Lê Cao.
Tenenhaus A. and Tenenhaus M., (2011), Regularized Generalized Canonical Correlation Analysis, Psychometrika, Vol. 76, Nr 2, pp 257-284.
Tenenhaus A., Phillipe C., Guillemot, V., Lê Cao K-A., Grill J., Frouin, V. Variable Selection For Generalized Canonical Correlation Analysis. 2013. (in revision)
wrapper.sgcca
, plotIndiv
, plotVar
, wrapper.rgcca
and http://www.mixOmics.org for more details.
data(nutrimouse) # need to unmap the Y factor diet if you pretend this is not a classification pb. # see also the function block.splsda for discriminant analysis where you dont # need to unmap Y. Y = unmap(nutrimouse$diet) data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid, Y = Y) # with this design, gene expression and lipids are connected to the diet factor # design = matrix(c(0,0,1, # 0,0,1, # 1,1,0), ncol = 3, nrow = 3, byrow = TRUE) # with this design, gene expression and lipids are connected to the diet factor # and gene expression and lipids are also connected design = matrix(c(0,1,1, 1,0,1, 1,1,0), ncol = 3, nrow = 3, byrow = TRUE) #note: the penalty parameters will need to be tuned wrap.result.sgcca = wrapper.sgcca(X = data, design = design, penalty = c(.3,.5, 1), ncomp = 2, scheme = "centroid") wrap.result.sgcca #did the algo converge? wrap.result.sgcca$crit # yes