Talk:Effective sample size
It would be helpful to put some reference to accompany "The case where the correlations are not uniform is somewhat more complicated". T 15:15, 15 September 2017 (UTC) — Preceding unsigned comment added by Taurvil (talk • contribs)
A reference to the first Neff calculation should be made as I cannot find it anywhere else I the literature and in fact I have found different formulations giving different answers for the same values. The more standard Nee=N(1-r)/(1+r) where r = lag 1 correlation. 97.104.197.234 (talk) 17:15, 23 December 2016 (UTC)
Should this article be merged with design effect?[edit]
Given that ESS is basically the inverse of design effect, I'm wondering if it makes more sense to merge it into design effect (specifically: make it a section there, and redirect this article to that section). What do others here think? Tal Galili (talk) 16:45, 28 May 2021 (UTC)
- (cc users who've edited this page in the last 50 edits and may want to weigh in: User:Beland, User:GünniX, User:Michael Hardy, User:Shellwood, User:Qwfp - you're all great wiki editors, sorry if the ping is too blunt. Just don't want to have this discussion drag for a year :) ) Tal Galili (talk) 20:42, 28 May 2021 (UTC)
- At a quick glance, it's unclear to me how they are related; both articles are too math-heavy to really be accessible to most readers. Some sort of visual explanation or non-equation example would help a lot. -- Beland (talk) 00:17, 29 May 2021 (UTC)
- Effective sample size = (1/(design effect)) - they have a simple linear relationship between the two. The fact that they are both difficult articles for someone not in the field, I agree (and I agree they should be improved). Tal Galili (talk) 07:24, 30 May 2021 (UTC)
- At a quick glance, it's unclear to me how they are related; both articles are too math-heavy to really be accessible to most readers. Some sort of visual explanation or non-equation example would help a lot. -- Beland (talk) 00:17, 29 May 2021 (UTC)