I have a moderate size (n about 650) data set with missing values. I created 20 fold imputed data sets using mice.
I used brm_multiple to estimate a logistic model (family bernoulli) with 19 predictors and a horseshoe prior on the imputed data sets.
The model ran without difficulty and appropriate ESS, Rhat, Pareto k estimates, etc were returned.
I called projpred:
fit.cvs ← cv_varsel(fit, method = ‘forward’, cv_method = ‘LOO’)
The cv_varsel estimation completed and the returned values are consistent with fit.object coefficient values.
The following warning appeared:
Warning: Using only the first imputed data set. Please interpret the results with caution until a more principled approach has been implemented.
I take this to imply that only the posteriors of the first imputed data set are submitted to cv_varsel.
I want to check the projpred estimated properties of the other imputed data sets.
I have looked at the brms, projpred, and posterior packages for an argument to call the posteriors of any other imputed data set. I am not able to identify how to do so.
I’d appreciate a pointer.
Nathan
PS My fit object has size 35089032 bytes. My fit.cvs object has size 911980392. It is about 25 times larger. I assume that this is expected.
sessionInfo()
R version 4.1.1 (2021-08-10)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 18.04.5 LTS
Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.7.1
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.7.1
brms_2.16.1
projpred_2.0.2
rstan_2.21.2
mice_3.13.0