linear.bayes: Bayesian QTL mapping via Linearized Likelihood

In bqtl: Bayesian QTL Mapping Toolkit

Bayesian QTL mapping via Linearized Likelihood

Description

The Bayesian QTL models via a likelihood that is linearized w.r.t. a fixed genetic model. By default, all one and two gene models (without epistasis) are fitted and a MCMC sampler is used to fit 3,4, and 5 gene and (optionally) larger models.

Usage

linear.bayes(x, ana.obj, partial=NULL, rparm, specs,
scope, subset, casewt, ...)

Arguments

x	a formula giving the QTL and the candidate loci or a varcov object
ana.obj	An analysis.object, see make.analysis.obj
partial	a formula giving covariates to be controlled
rparm	A ridge parameter. A value of 1 is suggested, but the default is 0.
specs	An optional list with components gene.number (to indicate the model sizes), burn.in (to indicate the number of initial MCMC cycles to discard), and n.cycles (to indicate how many MCMC cycles to perform for each model size). If no values are supplied, specs defaults to list(gene.number=c(1,2,3,4,5),burn.in=1,n.cycles=c(0,0,200,100,100))
scope	Not generally used. If supplied this will be passed to varcov.
subset	Not generally used. If supplied this will be passed to varcov.
casewt	Not generally used. If supplied this will be passed to varcov.
...	optional arguments to pass to twohk and swap

Details

This function is a wrapper for varcov, twohk, swap, and summary.swap, and a better understanding of optional arguments and the object generated is gained from their documentation.

Value

hk	The object returned by twohk
swaps	A list of objects returned by calls to swap. Element i in swaps is for i gene models.
smry	A list of objects returned by calls to summary.swap. Some elements may be NULL if no samples were requested or if the sampling process yielded degenerate results. Usually, this happens if no posterior is specified for the regression coefficients, i.e. if rparm=0 was used or implied
odds	A Vector of odds (relative to a no gene setup) for each model size evaluated. The odds are computed under a prior that places equal weights on models of each size considered (and are, therefore, Bayes Factors). If models of size 1 and 2 are not evaluated or if some degenerate results were encountered, this will be NULL
coefs	A vector of posterior means of the regression coefficients. If models of size 1 and 2 are not evaluated or if some degenerate results were encountered, this will be NULL
loc.posterior	A vector of locus-wise posterior probabilities that the interval covered by this locus contains a gene.If models of size 1 and 2 are not evaluated or if some degenerate results were encountered, this will be NULL
call	The call that generated this object

Author(s)

Charles C. Berry [e-mail]

References

Berry C.C.(1998) Computationally Efficient Bayesian QTL Mapping in Experimental Crosses. ASA Proceedings of the Biometrics Section. 164–169.

Examples

data( little.ana.bc )
little.lin <- linear.bayes( bc.phenotype~locus(all), little.ana.bc, rparm=1 )
par(mfrow=c(2,3))
plot( little.ana.bc, little.lin$loc.posterior, type="h" )
little.lin$odds
par(mfrow=c(1,1))
plot(fitted(little.lin), residuals(little.lin))

bqtl documentation built on Sept. 25, 2024, 1:08 a.m.