I recall earlier that in e.g. morph we were going to remove zero (<1e-8 I think) variance rows due to having pearson NaN. A user mentioned that doing so may skew results. We should look into whether that's the case and whether there's a better solution if so.
In GitLab by @timdiels on Oct 8, 2020, 11:02
I recall earlier that in e.g. morph we were going to remove zero (<1e-8 I think) variance rows due to having pearson NaN. A user mentioned that doing so may skew results. We should look into whether that's the case and whether there's a better solution if so.