New projpred 2.1.1 in CRAN

New projpred release in CRAN CRAN - Package projpred.

@fweber144 has done an awesome job by fixing many bugs, added new major features, and is the new main maintainer of the package (120 items in the release notes!)

projpred does Bayesian decision theoretical variable and model selection in case of large number of models with excellent performance and proper post-selection inference. projpred works especially smoothly with rstanarm and brms, but the reference model can be anything

The major new features are support for GAMs and GAMMs, linear vs smooth selection in GAMs, support for Student_t and Gamma families, option to use faster more approximate LOO, subsampling LOO, more parallelization (can fail in some platforms)

@fweber144 has been super efficient and as he is working 50% on projred (with my Academy of Finland grant funding) there will be more useful major new features appearing later this year. For some forthcoming features, we will announce here when the features are available in github as it’s useful to get extra user testing before submitting to CRAN


Thank you for this announcement and your kind words. I want to add that the former maintainer @AlejandroCatalina was still involved in this release, so thanks a lot to him, too!

Note that support for the Student-t and the Gamma family is still experimental; the Gamma family is even likely to run into an error. We hope to make them work soon.