pmf of betabinomial

[bgorissen]

It appears the term ln(n choose x) is missing from the log pmf calculation for the beta binomial distribution. You may want to rely on scipy.stats.betabinom.logpmf instead.

https://github.com/StatBiomed/BBMix/blob/master/bbmix/models/betabin_mix.py#L155C13-L155C25

            log_E_gammas[k] = np.log(pi) + gammaln(y + a) + gammaln(n - y + b) + gammaln(a + b) - \
                              (gammaln(a) + gammaln(b) + gammaln(n + a + b))

https://github.com/StatBiomed/BBMix/blob/master/bbmix/models/betabin_mix.py#L119

        return gammaln(y + a) + gammaln(n - y + b) + gammaln(a + b) - \
               (gammaln(a) + gammaln(b) + gammaln(n + a + b)) + np.log(pi)