krigr_cross_validation: Multithreaded cross validation routine for Omic Kriging.

Description

This is a flexible cross validation routine which wraps the Omic Kriging calculation. The user can specify the size of the test set, all the way to "Leave One Out" cross validation. Additionally, all relevant parameters in the okriging function are exposed. This function uses the doParallel package to distribute computation over multiple cores. If the phenotype is case/control, a ROCR AUC and GLM analysis is run and the results printed to screen.

Usage

krigr_cross_validation(cor.list, pheno.df, pheno.id = 1, h2.vec, covar.mat = NULL, nfold = 10, ncore = "all", verbose = FALSE, ...)

Arguments

cor.list

A list of correlation matrices used in Kriging. rownames and colnames of cor should be IID list and include idtest and idtrain.

pheno.df

A data frame with rownames set as sample IDs and a column containing phenotype data.

pheno.id

The name of the column in pheno which contains phenotype data to test.

h2.vec

has weights for each RM relatednes matrix

covar.mat

Data frame of covariates with rownames() set to sample IDs.

nfold

Select the number of cross validation rounds to run. The value "LOOCV" will run one round of cross validation for each sample in your dataset. The value "ncore" will set the test set size such that a single round runs on each core specified in the ncore option. Any numeric value will be set to the test size. Default runs 10 rounds of cross validation.

ncore

The number of cores available to distribute computaition across If a numeric value is supplied, that number of cores is registered. If the value "all" is supplied, all available cores are used.

verbose

Report rounds on cross validation on standard out.

...

Optional and unnamed arguments.

Value

A dataframe with three columns: sample ID, observed phenotype Ytest, and predicted phenotype Ypred