Applying the usual statistical methods after selecting variables is common in medical research but can lead to severe bias. Novel statistical methods for correct, post-selection inference have been implemented in the R package selcorr.
selcorr is an R package that makes correct statistical inference after variable selection.
selcorr version 1.0 can make post-selection inference for generalised linear models and some standard selection procedures, in particular for logistic or linear regression and for AIC (Akaike information criterion) selection. Confidence intervals and p-values for regression coefficients are corrected by parametric bootstrap calibration. Future versions of selcorr will be extended to other statistical models and selection procedures.
This tool should be used by statisticians for making correct statistical inference after variable selection.
selcorr’s code was developed by statisticians at the CTU Basel on behalf of the SCTO’s Statistics & Methodology Platform. It is one of the winning projects of the platform’s 2020 statistical programming grant.
The R statistics package selcorr: Post-Selection Inference for Generalized Linear Models is available on n the Comprehensive R Archive Network (CRAN). For detailed instructions on how to use this tool, please refer to the documentation on CRAN.
Feel free to contact us with any questions or feedback related to this resource!