This is particularly useful when the model is built with scale=T but pentrace does not apply that. In that case pentrace is applying the penalty to the unscaled coefficients to evaluate the log likelihood, but when the model is refit the penalty is applied to the scaled coefficients. With the changes in this branch you can now do
This is particularly useful when the model is built with scale=T but pentrace does not apply that. In that case pentrace is applying the penalty to the unscaled coefficients to evaluate the log likelihood, but when the model is refit the penalty is applied to the scaled coefficients. With the changes in this branch you can now do
pentrace(fit, scale=T)