search

sanssouci-org / sanssouci.python

Post hoc inference via multiple testing
GNU General Public License v3.0
6 stars 3 forks source link

Potential bug when using maxFP/curve_max_FP #30

Closed alexblnn closed 2 years ago

alexblnn commented 2 years ago

we need to be careful using the k_min idea (see https://github.com/pneuvial/sanssouci.python/pull/29) and computing bounds. If a sliced template (i.e. thr[k_min:]) is given as input to maxFP or curve_max_FP we give over optimistic bounds, since those functions assume that the templates starts at k=1 (see https://github.com/pneuvial/sanssouci.python/blob/main/sanssouci/post_hoc_bounds.py#L56)

pneuvial commented 2 years ago

As discussed a workaround seems to force thr[0:(k_min-1)] to be 0 instead of slicing the template. This should have the effect to make the k < k_min suboptimal in the calculation of the bound.

pneuvial commented 2 years ago

The k_min idea has been abandoned.