OmicKriging (version 1.0)

okriging: predict the phenotype of an individual using the remaining individuals

Description

Given "omics" correlation matrices, matrix weights, and covariates, okriging predicts the phenotype of an individual by using the phenotypes of the remaining individuals.

Usage

okriging(idtest, idtrain = NULL, corlist, H2vec, 
pheno, phenoname, Xcova = NULL)

Arguments

idtest

IID of individual(s) in the test set for phenotype prediction

idtrain

IID list of the remaining individuals in the training set for phenotype prediction

corlist

list of correlation matrices (rownames and colnames should be the IID list and include idtest and idtrain

H2vec

list same length as corlist of weights for each correlation matrix, sum should equal 1

pheno

data.frame of the phenotype (rownames should be the IID list and colnames should be: FID, IID, phenoname)

phenoname

name of the phenotype

Xcova

matrix of covariates with the IID list as the rownames

References

Wheeler HE, et al. (2013) Poly-Omic Prediction of Complex Traits: OmicKriging. arXiv:1303.1788 http://arxiv.org/abs/1303.1788

Cressie, N. (1993) Statistics for Spatial Data, Wiley, New York. p.154

Stein, M. L. (1999) Interpolation of Spatial Data: Some Theory for Kriging, Springer, New York.

See Also

readcorlist grm2mat

Examples

data(iddata)
data(corlist)
data(H2vec)
data(pheno)
data(phenoname)
data(Xcovamat)

idfamilylist=unique(iddata$FID)
pred = data.frame()
for(idtrio in idfamilylist){
  idtest = iddata$IID[iddata$FID ==idtrio]
  idtrain = iddata$IID[!(iddata$IID %in% idtest)]
  res = okriging(idtest,idtrain,corlist,H2vec,pheno,phenoname,Xcova = Xcovamat)
  pred = rbind(pred,res)
  print(res)
}
summary(lm(Ypred~Ytest,data=pred))