How many groups is enough?
Alain Zuur's response to a recent posting raises an interesting question. To use a random effects model what number of groups is actually sufficient? I have heard talk of a minimum of 20 groups but have seen numerous examples in books and published papers with much less than this. Is there a definitive reference on this?
Graham, Actually..it turned out that the data set for which the question was asked, had about 350 subjects I believe. But anyway....that is not your question. In general you see the magic "5" in some textbooks.....but for what it is worth...I recently had to program a ZIP for 2-way nested data in RBugs..and in order to do this, I started with 1-way and 2-way GLMMs (just to build up the code). And to check whether my code was "correct", I compared the results with that of 3-4 R packages (e.g. glmmPQL, lmer, glmml). The data set consisted of multiple observations per animal, for 5-30 animals per colony, and 9 colonies. I noticed that the estimated values for the variance for the random intercept colony differed a lot between these packages. But all came with similar estimates for the animal-within-colony random intercept. Not that it tells you that much (all packages giving the same result doesn't mean it is correct)....but it is a bit worrying. Perhaps a simulation study gives you a better answer. The data I use(d) are highly unbalanced..so that may have played a role as well. Alain
Dr. Alain F. Zuur First author of: 1. Analysing Ecological Data (2007). Zuur, AF, Ieno, EN and Smith, GM. Springer. 680 p. URL: www.springer.com/0-387-45967-7 2. Mixed effects models and extensions in ecology with R. (2009). Zuur, AF, Ieno, EN, Walker, N, Saveliev, AA, and Smith, GM. Springer. http://www.springer.com/life+sci/ecology/book/978-0-387-87457-9 3. A Beginner's Guide to R (2009). Zuur, AF, Ieno, EN, Meesters, EHWG. Springer http://www.springer.com/statistics/computational/book/978-0-387-93836-3 Other books: http://www.highstat.com/books.htm Statistical consultancy, courses, data analysis and software Highland Statistics Ltd. 6 Laverock road UK - AB41 6FN Newburgh Tel: 0044 1358 788177 Email: highstat at highstat.com URL: www.highstat.com URL: www.brodgar.com