@03rcooke has made me realise an unwelcome behaviour in bma(). One virtue of the model, in theory, is that it can account for zero counts. But actually we ask for the data to be provided on the log scale. So there's an inconsistency that we've not really documented.
@03rcooke has made me realise an unwelcome behaviour in
bma(). One virtue of the model, in theory, is that it can account for zero counts. But actually we ask for the data to be provided on the log scale. So there's an inconsistency that we've not really documented.