predict.basglm not compatible with MPM estimator

[aagaardk]

When estimator = 'MPM' predict.basglm produces an error regarding the prior values.

Reproducible example:

data(Pima.tr, package="MASS")
data(Pima.te, package="MASS")
Pima.bas = bas.glm(type ~ ., data=Pima.tr, n.models= 2^7, method="BAS",
                   betaprior=CCH(a=1, b=nrow(Pima.tr)/2, s=0), family=binomial(),
                   modelprior=uniform())
pred = predict(Pima.bas, newdata=Pima.te, estimator = 'MPM')

#Error in bas.lm(eval(object$call$formula), data = eval(object$call$data,  : 
#  prior  1 is not one of  g-prior, hyper-g, hyper-g-laplace, hyper-g-n, AIC, BIC, ZS-null, ZS-full, EB-local, EB-global, JZSprior

Desktop (please complete the following information):

• OS: Windows x86_64-w64-mingw32/x64 (64-bit)
• R Version 4.0.2

[merliseclyde]

For MPM, predict extracts the median probability model and runs bas a second time to create a new bas object. Currently predict code calls bas.lm, rather than bas.glm.

Fix: add a switch to check class of object before calling bas.lm or bas.glm