Effect size from post-hoc tests

[esiotort]

Hello!

Is there any way to calculate generalised eta-squared for contrasts run on afex anova's using the lsmeans/multcomp procedures detailed in your post-hoc tests vignette?

Cheers

[singmann]

Unfortunately not at the moment (and I do not think anytime soon). The reason is that those types of standardized effect sizes require sums of squares which are not reported by lsmeans or multcomp.

Instead, I would simply report the estimate from the lsemans output as effect. If you necessary need a standardized effect size (which might not be the best thing in the world) an effect size based on the output from lsmeans (i.e., estimate and SE) may be possible, but this may depends on your design (within- or between-factors) and what you believe is a good standardized effect size. To state this clearly, I am not a big fan of standardized effect sizes.

PS: I will leave this open for future reference and people looking for the same thing.

[esiotort]

Thanks for the answer

[singmann]

See also a somewhat longer response at: https://afex.singmann.science/forums/topic/compute-effect-sizes-for-mixed-objects/#post-295