bqtl {bqtl} | R Documentation |
Bayesian QTL Model Fitting
Description
Find maximum likelihood estimate(s) or posterior mode(s) for QTL model(s). Use Laplace approximation to determine the posterior mass associated with the model(s).
Usage
bqtl(reg.formula, ana.obj, scope = ana.obj$reg.names, expand.specials = NULL, ...)
Arguments
reg.formula |
A formula.object like |
ana.obj |
The result of |
scope |
passed to |
expand.specials |
passed to |
... |
Arguments to pass to |
Details
This function is a wrapper for lapadj
. It does a lot
of useful packaging through the configs
terms. If there
is no configs
term, then the result is simply the output of
lapadj
with the call
attribute replaced by the
call to bqtl
Value
The result(s) of calling lapadj
.
If configs
is used in the reg.formula
, then the
result is a list with one element for each formula. Each element is the
value returned by lapadj
Author(s)
Charles C. Berry cberry@ucsd.edu
References
Tierney L. and Kadane J.B. (1986) Accurate Approximations for Posterior Moments and Marginal Densities. JASA, 81,82–86.
See Also
Examples
data(little.ana.bc ) # load BC1 dataset
loglik( bqtl( bc.phenotype ~ 1, little.ana.bc ) ) #null loglikelihood
#on chr 1 near cM 25
loglik(bqtl(bc.phenotype~locus(chromo=1,cM=25),little.ana.bc))
little.bqtl <- # two genes with epistasis
bqtl(bc.phenotype ~ m.12 * m.24, little.ana.bc)
summary(little.bqtl)
several.epi <- # 20 epistatic models
bqtl( bc.phenotype ~ m.12 * locus(31:50), little.ana.bc)
several.main <- # main effects only
bqtl( bc.phenotype ~ m.12 + locus(31:50), little.ana.bc)
max.loglik <- max( loglik(several.epi) - loglik(several.main) )
round(
c( Chi.Square=2*max.loglik, df=1, p.value=1-pchisq(2*max.loglik,1))
,2)
five.gene <- ## a five gene model
bqtl( bc.phenotype ~ locus( 12, 32, 44, 22, 76 ), little.ana.bc , return.hess=TRUE )
regr.coef.table <- summary(five.gene)$coefficients
round( regr.coef.table[,"Value"] + # coefs inside 95% CI
qnorm(0.025) * regr.coef.table[,"Std.Err"] %o%
c("Lower CI"=1,"Estimate"=0,"Upper CI"=-1),3)