createXXblocks {multiridge}

Creates list of (unscaled) sample covariance matrices

Description

Creates list of (unscaled) sample covariance matrices X_b %*% t(X_b) for data blocks b = 1,..., B.

Usage

createXXblocks(datablocks, datablocksnew = NULL, which2pair = NULL)

Arguments

datablocks

List of data frames or matrices

datablocksnew

List of data frames or matrices

which2pair

Integer vector of size 2 (or NULL)

Details

The efficiency of multiridge for high-dimendional data relies largely on this function: all iterative calculation are performed on the out put of this function, which contains B blocks of nxn matrices. If which2pair != NULL, the function adds a paired covariance block, pairing the two data blocks corresponding to the elements of which2pair. If predictions for new samples are desired, one also needs to specify datablocksnew, which should have he exact same format as datablocks with matching column dimension (number of variables).

Value

List. Same number of component as datablocks when which2pair = NULL, or augmented with one paired data block. Dimension is nxn for all components.

See Also

createXblocks, which is required when parameter estimates are desired (not needed for prediction). A full demo and data are available from:
https://drive.google.com/open?id=1NUfeOtN8-KZ8A2HZzveG506nBwgW64e4

Examples

#Example
#Simulate
Xbl1 <- matrix(rnorm(1000),nrow=10)
Xbl2 <- matrix(rnorm(2000),nrow=10)

#check whether dimensions are correct
ncol(Xbl1)==nrow(Xbl2)

#create cross-product
XXbl <- createXXblocks(list(Xbl1,Xbl2))

#suppose penalties for two data types equal 5,10, respectively
Sigma <- SigmaFromBlocks(XXbl,c(5,10))

#check dimensions (should be n x n)
dim(Sigma)