BayesFactor: Bayesian model selection via Bayes factors.

In fboehm/qtlbim: QTL Bayesian Interval Mapping

Description

Model-averaged posteriors and Bayes factors computed for number and pattern of QTL, chromosomes and pairs of chromosomes showing epistasis.

Usage

qb.bf(...)
qb.BayesFactor(qbObject, items = c("nqtl","pattern","chrom","pairs"),
  cutoff.pattern, cutoff.pairs = 1, nmax = 15, epistasis = TRUE, ...)
## S3 method for class 'qb.BayesFactor'
plot(x, ...)
## S3 method for class 'qb.BayesFactor'
summary(object, sort = TRUE, digits = 3, ...)
## S3 method for class 'qb.BayesFactor'
print(x, ...)

Arguments

qbObject	An object of class qb.
object	Object of class qb.BayesFactor.
x	Object of class qb.BayesFactor.
items	Items to include in model selection assessment. These are nqtl = Number of QTLs; pattern = Pattern of QTL across chromosomes as comma-separated chromosome numbers and colon-separated chromosome pairs; chrom = Chromosome; pairs = Epistatic pairs of chromosomes.
cutoff.pattern	Percent cutoff for pattern inclusion in model selection. Default is 0.25 (0.5) if epistasis is TRUE (FALSE).
cutoff.pairs	Percent cutoff for epistatic pair inclusion in model selection.
nmax	Maximum number of model terms included per item (for items "pattern" and "pairs" only).
epistasis	Include epistasis in patterns if TRUE.
sort	Sort by Bayes factor if TRUE.
digits	Number of significant digits for summary.
...	Additional arguments passed to generic plot, summary or print.

Details

qb.BayesFactor (or qb.bf for short) creates model selection results for selected items. These are based on marginal posteriors and priors, averaged over all other model parameters. The posterior may be influenced by prior, while Bayes factors are empirically less sensitive for QTL model selection. The Bayes factors are computed relative to the smallest term for each item, using the ratios of posterior/prior. Any pair of model terms can be compared as the ratio of their Bayes factors.

Value

List with items, each containing:

posterior	Posterior frequency of MCMC samples.
prior	Prior frequency.
bf	Rank-ordered Bayes factors relative to smallest value.
bfse	Approximate standard error for bf computed using binomial variance of MCMC samples.

Author(s)

Brian S. Yandell, [e-mail]

References

http://www.qtlbim.org

See Also

plot.qb, qb.mcmc

Examples

data(qbExample)

temp <- qb.BayesFactor(qbExample)
summary(temp)
plot(temp)

fboehm/qtlbim documentation built on Feb. 16, 2021, 12:04 a.m.