coef throws warnings with bas.glm objects if prior is AIC or BIC

[merliseclyde]

Describe the bug
priors in bas.glm are objects, while in bas.lm character strings are used to refer to the prior, so the check to see if the prior is one of AIC or BIC in the coef function fails as it is comparing to object$prior which completes to object$priorprobs and the wrong length. No warning occurs if there is only a single model as when restricting the number of models.

To Reproduce
Steps to reproduce the behavior:

 pima.bas = bas.glm(formula = type ~ bp + bmi, data = Pima.tr, 
                     family = binomial(), 
                     betaprior = bic.prior(), 
                     modelprior = uniform(), 
                     method = "BAS")

 coef(pima.bas)

Expected behavior
no warnings or errors

** Offending Code **

Starting line 126

  if (!(object$prior == "AIC" || object$prior == "BIC")) {
    conditionalsd[, -1] <- sweep(conditionalsd[, -1, drop = F], 1,
      sqrt(shrinkage),
      FUN = "*"
    )
  }

and line 140

      df <- df - object$rank
    } else {
      df <- df - 1

Desktop (please complete the following information):

• OSX
• R Version 4.2.2

Additional context
check if this impacts predictions too

[merliseclyde]

Fixed in main version 1.6.5