OmicKriging (version 1.0)
OmicKriging for Phenotypic Prediction
Description
This package provides functions to generate correlation
matrices from SNP, gene expression, methylation or other
'omics' datasets and to use these matrices to predict the
phenotype of an individual by using the phenotypes of the
remaining individuals through kriging. Kriging is a
geostatistical method for optimal prediction or best unbiased
linear prediction. It consists of predicting the value of a
variable at an unobserved location as a weighted sum of the
variable at observed locations. Intuitively, it works as a
reverse linear regression: instead of computing correlation
(univariate regression coefficients are simply scaled
correlation) between a dependent variable Y and independent
variables X, it uses known correlation between X and Y to
predict Y.