non normal distribution error thrown when using ANCOVA

[thebeancounter]

I am trying to use ANCOVA to find a relation between two variables eliminating one other variable. All of them are numeric. I am getting an error thrown saying the predicted variable is not normally distributed, no visible outliers, no NA values, and the same data works well with partial_corr.

Where can I check what is the data caused that error?

[raphaelvallat]

Hi @thebeancounter,

Can you copy-paste the full error message and provide minimal code example to reproduce the error? There's no such error message in Pingouin, so the error might come from the lower-level statsmodels function.