contingencyTableBF: Function for Bayesian analysis of one- and two-sample designs

In BayesFactor: Computation of Bayes Factors for Common Designs

Function for Bayesian analysis of one- and two-sample designs

Description

This function computes Bayes factors for contingency tables.

Usage

contingencyTableBF(
  x,
  sampleType,
  fixedMargin = NULL,
  priorConcentration = 1,
  posterior = FALSE,
  callback = function(...) as.integer(0),
  ...
)

Arguments

x	an m by n matrix of counts (integers m,n > 1)
sampleType	the sampling plan (see details)
fixedMargin	for the independent multinomial sampling plan, which margin is fixed ("rows" or "cols")
priorConcentration	prior concentration parameter, set to 1 by default (see details)
posterior	if TRUE, return samples from the posterior instead of Bayes factor
callback	callback function for third-party interfaces
...	further arguments to be passed to or from methods.

Details

The Bayes factor provided by contingencyTableBF tests the independence assumption in contingency tables under various sampling plans, each of which is described below. See Gunel and Dickey (1974) for more details.

For sampleType="poisson", the sampling plan is assumed to be one in which observations occur as a poisson process with an overall rate, and then assignment to particular factor levels occurs with fixed probability. Under the null hypothesis, the assignments to the two factors are independent. Importantly, the total N is not fixed.

For sampleType="jointMulti" (joint multinomial), the sampling plan is assumed to be one in which the total N is fixed, and observations are assigned to cells with fixed probability. Under the null hypothesis, the assignments to the two factors are independent.

For sampleType="indepMulti" (independent multinomial), the sampling plan is assumed to be one in which row or column totals are fixed, and each row or column is assumed to be multinomially distributed. Under the null hypothesis, each row or column is assumed to have the same multinomial probabilities. The fixed margin must be given by the fixedMargin argument.

For sampleType="hypergeom" (hypergeometric), the sampling plan is assumed to be one in which both the row and column totals are fixed. Under the null hypothesis, the cell counts are assumed to be governed by the hypergeometric distribution.

For all models, the argument priorConcentration indexes the expected deviation from the null hypothesis under the alternative, and corresponds to Gunel and Dickey's (1974) "a" parameter.

Value

If posterior is FALSE, an object of class BFBayesFactor containing the computed model comparisons is returned.

If posterior is TRUE, an object of class BFmcmc, containing MCMC samples from the posterior is returned.

Note

Posterior sampling for the hypergeometric model under the alternative has not yet been implemented.

Author(s)

Richard D. Morey ([e-mail])

Tahira Jamil ([e-mail])

References

Gunel, E. and Dickey, J., (1974) Bayes Factors for Independence in Contingency Tables. Biometrika, 61, 545-557

Examples

## Hraba and Grant (1970) doll race data
data(raceDolls)

## Compute Bayes factor for independent binomial design, with
## columns as the fixed margin
bf = contingencyTableBF(raceDolls, sampleType = "indepMulti", fixedMargin = "cols")
bf

## Posterior distribution of difference in probabilities, under alternative
chains = posterior(bf, iterations = 10000)
sameRaceGivenWhite = chains[,"pi[1,1]"] / chains[,"pi[*,1]"]
sameRaceGivenBlack = chains[,"pi[1,2]"] / chains[,"pi[*,2]"]
hist(sameRaceGivenWhite - sameRaceGivenBlack, xlab = "Probability increase",
  main = "Increase in probability of child picking\nsame race doll (white - black)",
  freq=FALSE, yaxt='n')
box()

BayesFactor documentation built on Sept. 22, 2023, 1:06 a.m.