search

0todd0000 / spm1d

One-Dimensional Statistical Parametric Mapping in Python
GNU General Public License v3.0
61 stars 21 forks source link

Effect sizes for Hotelling's test? #277

Closed The-DanKadlec closed 5 months ago

The-DanKadlec commented 7 months ago

Dear Todd,

I'm running a 2 sample Hotelling's test to evaluate the difference in GRF (Fz and Fy) between 2 cohorts, which is straight forward. Now, is it possible to calculate an effect size to help the interpretation of the T2 statistic? I read about Partial eta squared and Multivariate Cohen's d based on the Mahalanobis distance as theoretical possibilities for multivariate data, however, I haven't found much in-depth information on either (assumptions, confidence intervals, thresholds, code examples).

Can you provide any guidance if any of these is a feasible approach to help visualise and interpret the data? Or if there generally are or are not ways to add ES for this dataset?

Many thanks for your support, Dan

0todd0000 commented 7 months ago

Yes, multivariate effect sizes can indeed be calculated, but they are unfortunately not yet supported in spm1d. The test statistic values (in this case: $T^2$-values) exactly follow effect size value patterns across the 1D domain, so I think it may be fine to report the $T^2$-values in lieu of effect size.