Within strata variance

It is important that the variance of the mean is independent of the variance among strata and depends only on the variance within strata and on the number of determinations. 

Suppose that the bulk of material to be analyzed is sampled by taking n strata, each of which provides M2 samples, and that Mj determinations are to be carried out on each sample. The various strata are assumed to be equal in size and in variance within strata. 

Since the smaller is reciprocal of the variance, the larger is the variance, and vice versa, the consequence is that trials for which the treatment estimate has a smaller variance should be given more weight. Suppose we have a series of perfectly balanced trials, so that for trial i we have rij patients in the experimental and fij in the control group. If we perform a meta-analysis of continuous outcomes using original data and the standard linear model, then we conventionally assume that variances within strata are equal. Where that is so, the variance of the Ith treatment estimate is... 

In stratified sampling the population to be examined is first divided into a number of portions called strata, followed by random sampling within each stratum. The overall variance, is... 

---