Mixed 2-way ANOVA

[dokato]

My RM ANOVA works fine:

pg.rm_anova(data = df, dv='time', subject='pid', within=['axis', 'congr'])

I thought it would be easy to extrapolate to mixed ANOVA by introducing another between subject variable. But this:

pg.mixed_anova(data=df, dv='time', within=['axis', 'congr'], between='shift', subject='pid')

errors: 'DataFrame' object has no attribute 'dtype'

Is there a plans to introduce 2-way mixed ANOVA, eg. for a design 2x2x2?

[raphaelvallat]

Hi @dokato,

For now, only one-within / one-between factors are supported in mixed_anova. It would actually require a lot of changes to the code to add support for multiple between or within factor, so I doubt this will be implemented in the near future.

For more complex ANOVA designs, I strongly recommend using JASP.

Thanks!
Raphael

[dokato]

@raphaelvallat ah pity, thanks for adding it to the roadmap though ;)

[sappelhoff]

based on how often this question comes up (also on gitter), it may be worthwhile to add a short check for mixed_anova: if more than one within or between factor are specified, a ValueError is raised, saying that this is not yet implemented, and pointing to this issue, asking for somebody to implement it :-)

[raphaelvallat]

Good idea @sappelhoff! Please feel free to submit it as a PR, or else I will do it for the next stable release.