In Wallisch et al 2021, dependent on the post-selection investigation measures (e.g. RCB, VIF, etc.), either the non-parametric bootstrap or subsampling with 0.5 as sampling proportion is recommended. It would be really nice, if type.resampling = "Wallisch_2021" (as an additional option) in abe.resampling would compute both resampling methods and then use the correct resampling method for each measure. The current implementation is prone to errors.
biometrician
commented
1 year ago
In Wallisch et al 2021, dependent on the post-selection investigation measures (e.g. RCB, VIF, etc.), either the non-parametric bootstrap or subsampling with 0.5 as sampling proportion is recommended. It would be really nice, if
type.resampling = "Wallisch_2021"(as an additional option) inabe.resamplingwould compute both resampling methods and then use the correct resampling method for each measure. The current implementation is prone to errors.