When a model is not full rank, and hence R2 is not defined using model fitting without pivoting, calculation of the log marginals under the ZS prior lead to errors, rather than returning NA's and a warning.
May be a problem with other priors as well, such as the hyper-g family
When a model is not full rank, and hence R2 is not defined using model fitting without pivoting, calculation of the log marginals under the ZS prior lead to errors, rather than returning NA's and a warning.
May be a problem with other priors as well, such as the hyper-g family