I want to compute Bayes Factor for one-sample proportion test where the null hypothesis is that-
The proportions for the levels for the nominal variable are not different from the theoretical proportions
I know how to use proportionBF from BayesFactor package to do this when there are two proportions with the null hypothesis being that the two levels are equally likely (p = 0.5). This is equivalent to stats::binom.test.
# frequentist test (p-value > 0.05)
broom::tidy(stats::binom.test(x = 21, n = 44, p = 0.5))
#> # A tibble: 1 x 8
#> estimate statistic p.value parameter conf.low conf.high method
#> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
#> 1 0.477 21 0.880 44 0.325 0.633 Exact~
#> # ... with 1 more variable: alternative <chr>
# quantifying evidence in favor of null hypothesis
library(BayesFactor)
1/ proportionBF(y = 21, N = 44, p = 0.5, rscale = 0.707)
#> Bayes factor analysis
#> --------------
#> [1] Null, p=0.5 : 3.724132 ±0.02%
#>
#> Against denominator:
#> Alternative, p0 = 0.5, r = 0.707, p =/= p0
#> ---
#> Bayes factor type: BFproportion, logistic
But I don't know how to do this analysis when there are more than two proportions, the equivalent of stats::chisq.test (a goodness of fit test).
# frequentist test (p-value > 0.05)
table(mtcars$cyl)
#>
#> 4 6 8
#> 11 7 14
broom::tidy(stats::chisq.test(x = table(mtcars$cyl)))
#> # A tibble: 1 x 4
#> statistic p.value parameter method
#> <dbl> <dbl> <dbl> <chr>
#> 1 2.31 0.315 2 Chi-squared test for given probabilities
How can I compute Bayes Factor for this test using BayesFactor package?
(P.S. If not possible, I'd also be interested in knowing any other package that can do the same.)
Author of BayesFactor answered this question here:
https://gist.github.com/richarddmorey/a4cd3a2051f373db917550d67131dba4