cvselpscca: Cross validation for Sparse Canonical Correlation Analysis

In lasandrall/SELPCCA: Sparse Canonical Correlation Analysis for Associating Mutiple High Dimensional Data

Description

Peforms nfolds cross validation to select optimal tuning parameters for SELPCCA based on training data. If you want to apply optimal tuning parameters to testing data, you may also use multiplescca.

Usage

cvselpscca(Xdata1=Xdata1,Xdata2=Xdata2,ncancorr=ncancorr,CovStructure="Iden",
          isParallel=TRUE,ncores=NULL,nfolds=5,ngrid=10,
          standardize=TRUE,thresh=0.0001,maxiteration=20)

Arguments

Xdata1	A matrix of size n \times p for first dataset. Rows are samples and columns are variables.
Xdata2	A matrix of size n \times q for second dataset. Rows are samples and columns are variables.
ncancorr	Number of canonical correlation vectors. Default is 1.
CovStructure	Covariance structure to use in estimating sparse canonical correlation vectors. Either "Iden" or "Ridge". Iden assumes the covariance matrix for each dataset is identity. Ridge uses the sample covariance for each dataset. See reference article for more details.
isParallel	TRUE or FALSE for parallel computing. Default is TRUE.
ncores	Number of cores to be used for parallel computing. Only used if isParallel=TRUE. If isParallel=TRUE and ncores=NULL, defaults to half the size of the number of system cores.
nfolds	Number of cross validation folds. Default is 5.
ngrid	Number of grid points for tuning parameters. Default is 10 for each dataset.
standardize	TRUE or FALSE. If TRUE, data will be normalized to have mean zero and variance one for each variable. Default is TRUE.
maxiteration	Maximum iteration for the algorithm if not converged. Default is 20.
thresh	Threshold for convergence. Default is 0.0001.

Details

The function will return several R objects, which can be assigned to a variable. To see the results, use the “$" operator.

Value

hatalpha	Estimated sparse canonical correlation vectors for first dataset.
hatbeta	Estimated sparse canonical correlation vectors for second dataset.
CovStructure	Covariance structure used in estimating sparse canonical correlation vectors. Either "Iden" or "Ridge".
optTau	Optimal tuning parameters for each dataset.
maxcorr	Estimated canonical correlation coefficient.
tunerange	Grid values for each dataset used for searching optimal tuning paramters.

References

Sandra E. Safo, Jeongyoun Ahn, Yongho Jeon, and Sungkyu Jung (2018) , Sparse Generalized Eigenvalue Problem with Application to Canonical Correlation Analysis for Integrative Analysis of Methylation and Gene Expression Data. Biometrics

See Also

multiplescca

Examples

library(SELPCCA)
##---- read in data
data(DataExample)

Xdata1=DataExample[[1]]
Xdata2=DataExample[[2]]


##---- call cross validation to estimate first canonical correlation vectors
ncancorr=1
mycv=cvselpscca(Xdata1=Xdata1,Xdata2=Xdata2,ncancorr=ncancorr,CovStructure="Iden",
                isParallel=FALSE,ncores=NULL,nfolds=5,ngrid=10,
                standardize=TRUE,thresh=0.0001,maxiteration=20)

#check output
train.correlation=mycv$maxcorr

optTau=mycv$optTau

hatalpha=mycv$hatalpha

hatbeta=mycv$hatbeta

#obtain correlation plot using training data
scoresX1=Xdata1%*% hatalpha
scoresX2=Xdata2%*% hatbeta
plot(scoresX1, scoresX2,lwd=3,
       ,xlab=paste(
         "First Canonical correlation variate for dataset", 1),
         ylab=paste("First Canonical correlation variate for dataset", 2),
         main=paste("Correlation plot for datasets",1, "and" ,2, ",", "\u03C1 =", mycv$maxcorr))


#obtain correlation plot using testing data

Xtestdata1=DataExample[[3]]
Xtestdata2=DataExample[[4]]
scoresX1=Xtestdata1%*%hatalpha
scoresX2=Xtestdata2%*%hatbeta
mytestcorr=round(abs(cor(Xtestdata1%*%hatalpha,Xtestdata2%*%hatbeta)),3)

plot(scoresX1, scoresX2,lwd=3,xlab=paste(
         "First Canonical correlation variate for dataset", 1),
         ylab=paste("First Canonical correlation variate for dataset", 2),
         main=paste("Correlation plot for datasets",1, "and" ,2, ",", "\u03C1 =", mytestcorr))

lasandrall/SELPCCA documentation built on June 8, 2020, 12:38 a.m.