Accounting for measurement error during variable selection with projpred (possibly with rstanarm or brms?)

Hi @dafanast,

I’m also trying to replicate the SPCA-based reference model approach in Pavone (2023) – given that the formula definition in brms appears to struggle with high numbers of predictors – and wondered how you managed to get this to work with later version(s) of projpred, as the paper’s GitHub code doesn’t work as is. May I ask how you wrote the extract_model_data function to work with the SPCA reference model?

Thanks in advance!