I have multiple characteristics of a population available (gender, age, place), and I have the answer for each individual for a particular yes-no question available done previously. For example, here is a sample population. pop = data.frame(y=c(1,1,0,1,0,1,0,1,0,0,0), gender=c(1,1,0,1,0,0,1,1,0,0,0), age=c(21,20,45,32,20,19,33,20,34,35,22), place=c(1,3,2,4,1,3,5,2,1,5,2)) I want to ask the same question to the same population again. The answer may have changed, but there is a reasonable expectation it may not have. What is the way to decide which of these characteristics (single or in combination) is best to use as a strata in stratified sampling to reduce the error? More importantly, how do I justify my choice? (Can I base it on some statistics of the previous)? I understand from [this](http://stats.stackexchange.com/questions/185229/when-should-you-choose-stratified-sampling-over-random-sampling) question answers that running an `anova` for each strata might help. > anova(lm(y~place+gender+age, data=pop)) Analysis of Variance Table Response: y Df Sum Sq Mean Sq F value Pr(>F) place 1 0.00147 0.00147 0.0086 0.92852 gender 1 1.17581 1.17581 6.9329 0.03376 * age 1 0.36281 0.36281 2.1392 0.18697 Residuals 7 1.18719 0.16960 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 So I should probably use `gender` alone as the strata here. Others are not of much use. Is this the correct way?