bqtl.fitter: Get loglikelihoods for many models of a common form

Description

For a single type of model, this function evaluates multiple models that differ only in terms of the loci involved. The looping is all done by internal C functions, so this is faster than simply using bqtl to do the same thing.

Usage

bqtl.fitter(setup, loc.mat, ana.obj)

Arguments

setup

The object returned by bqtl( <...> , setup=TRUE )

loc.mat

A matrix of locus numbers, s.t. nrow(loc.mat) equals the number of loci in setup

ana.obj

An analysis.object. Usually the one used in setup

Value

For now it only returns the loglikelihood. But it would be trivial to build an option that would allow other quantities computed to be returned, and this should probably be done. However, some care is needed to keep objects from becoming unmanageably large.

Details

In order to avoid the computational overhead of running large loops of very repetitive operations in R/S, bqtl.fitter used after the setup=TRUE option in bqtl will loop through the loci specified in loc.mat using internal C code. This is many times faster than running the same code via bqtl.

Examples

Run this code

data( little.ana.bc )
little.setup <-
bqtl( bc.phenotype~locus(1)*locus(2), little.ana.bc, setup=TRUE )
combos <- t( as.matrix( expand.grid( 1:21, 44:64 ) ) )
little.update <- bqtl.fitter(little.setup, combos, little.ana.bc)
little.res <- matrix( little.update, nr=21 )
image( 1:21, 44:64, little.res )
rm(little.ana.bc, little.update, little.res  )