Another question on the same subject, but this time somewhat unrelated to GrowthBook itself. If I want to analyze users by language (8 segments) and by device (3 segments: mobile, tablet, desktop), should I make a correction for overall 11 segments or should I correct separately for device (3 comparisons) and language (8 comparisons). I feel intuitively that the correction should be made for 11 overall segments. I am assuming Growthbook makes correction within the attribute and not across all values of the attributes?

Growthbook will only correct within one set of dimensional results at a time. We don't really have a good way right now to correct across multiple, but this might change as we let users define a family of "key" dimensions, a feature we're exploring now.

That’s what I thought exactly: just specify the primary attributes you may want to explore during the experiment setup so that the correction applies across specified segments. Like, if I add language and device to the primary attributes, the correction applies across combined number of values in those attributes. If I have 8 languages and 3 devices, that would be 11 segments and 11 p-values that need to be corrected for multiple comparison, if my stats understanding is right