I know there is the event per variable ratio of 10 to 1 for a regression analysis, where you should have ten event for every predictor in your model.
I was wondering what the rules are for a complex survey design. Is it still the same, or are they new or additional sets of rules for the number of predictors in an analysis.
The correct calculation of required sample sizes for complex surveys is a good few chapters of a book so I won't try to recreate it here other than to say that you are indeed correct that you need to include your survey design parameters (strata, weights, etc) in the calculation of your sample size.
The 10 to 1 rule is a more of a rule of thumb that seems to have persisted even though the support for it is mixed at best. I would suggest taking a look at Holtman et al. and then finding a good survey analysis book that is used in 3rd year or postgrad statistics subjects like Complex Surveys by Lumley.
