p-GG-corr for Interaction in mixed anova procedure retruns NaN

[MashRoofa]

Hello, im just curious if the result i get is normal or due to an error. My data is provided in a long format to the procedure and i ensured that there are no NaN values in the dependent variable.

I have one within factor (VIS_TYPE) with 6 categories and one between factor (WM) with 3 categories.

Sphericity was not assumed for the within factor categories therefore i wanted to interpret the Greenhouse-Geisser corrected p-value but it returns NaN as seen in the attached image.

Help is appreciated! Thank you!

[raphaelvallat]

Hi @MashRoofa,

The output is expected. The sphericity and Greenhouse-Geisser correction are only calculated for the within-subject factor (VIS_TYPE). It is thus expected that Pingouin returns NaN for the between-subject factor (WM) and interaction. In your case, you can see tha the p-GG-corr of the within-subject factor is <0.05 so the main within-factor effect is still significant after correction.

Hope this helps, Raphael

[MashRoofa]

Hi @raphaelvallat,

I appreciate the reply.

Im fairly new to statistics so please correct me if Im wrong.

Currently Im following a tutorial on SPSS (Tutorial offered by Laerd Statistics) and Im trying to apply what Im learning on Python, since the current project I'm working on requires that Python is used. Below (see attached image) you can see one part of the results captured when performing Two WAY Mixed ANOVA on SPSS.

As you can see in the row of the interaction results (middle row) SPSS provides the corrected p-value (i.e., Greenhouse-Geisser correction) and this is a specific corrected p-value for the interaction (i.e., not just for within factor). Moreover, this correction is also provided for the within factor as you mention above. So my question is, how can i get the corrected p-value for the interaction in pingouin?

Thank you!

[raphaelvallat]

You're not wrong — unfortunately the GG correction for the interaction is not implemented in Pingouin at this stage. I would therefore recommend sticking to SPSS or JASP for a free alternative.

Thanks! Raphael

[MashRoofa]

I understand, thank you for your replies! Are you aware of any python implementations of Two WAY Mixed ANOVA that also return the correction for the interaction?

Also since pingouin is such a great tool are there any plans to have this as a feature in the near future?

[raphaelvallat]

I don't know any Python implementation. I'd like to implement it but my time is very short right now and I don't anticipate it'll get better in the next months. This is unlikely to be implemented within the next 6 months.

Thanks, Raphael

[MashRoofa]

I understand friend, thank you ☺️