add a case for new versions of scipy (>1.16.3) where a new stats.mannwhitneyu parameter "method" defaults to "auto". auto uses "asymptotic" for most cases except where one group has fewer than 8 samples, in which case it uses "exact". This both changes the result from the original behavior and can be prohibitively slow when the other group is large (e.g. negative controls)
mhorlbeck
commented
1 year ago
add a case for new versions of scipy (>1.16.3) where a new stats.mannwhitneyu parameter "method" defaults to "auto". auto uses "asymptotic" for most cases except where one group has fewer than 8 samples, in which case it uses "exact". This both changes the result from the original behavior and can be prohibitively slow when the other group is large (e.g. negative controls)