Closed medsoccer closed 12 months ago
we've experimented with adding imputation to MS-DAP, but in the benchmark datasets that we evaluated the additional imputation step caused the ROC performance to decrease (in analyses analogous to those shown in the MS-DAP paper). We've explored various imputation algorithms, from KNN/simple to random forests. So after making these observations we didn't invest any time in bringing imputation into the pipeline.
I think the imputation approach described in the latest MSstats manuscript is interesting and theoretically sound, however it was not included in our previous assessment of imputation and we haven't put in the work of incorporating it into MS-DAP so we can benchmark all DEA algorithms with/without that specific imputation approach (probably not trivial, no time to explore this atm).
love the package and it's power.
just wondering if there's a way of adding an imputation step into the pipeline?