Today on Andrew Gelman’s blog he discusses whether it is better to use Stata, SAS, or R to run multi-level models on large datasets. Since I am in the process of (possibly) creating a short course on R and since I know very little about Stata and SAS, the topic was definitely of interest. As Gelman often does, there is no distinct correct answer (I like this btw….too many people believe they know the correct answer). Basically he says that if the data in all sub-groups is large, then analysis can be performed separately for all subgroups. In this case, any of the above software packages might work fine. But if an analysis needs to use the entire dataset, then he suggests that Stata > SAS >> R (Stata is better than SAS, but both are much better than R).