I'm currently using the Pooled Variance method, but in my case I cannot assume that every population variance is the same.
Is there a method for these cases ?
I'm currently using the Pooled Variance method, but in my case I cannot assume that every population variance is the same.
Is there a method for these cases ?
Copyright © 2021 JogjaFile Inc.
Here's what I would try. Find estimates of $E[X^2]$ for each group and then an overall estimate using a weighting scheme by group count. Do the same for the mean. Then estimate the overall variance with $$\hat \sigma^2=E[X^2]-(E[X])^2.$$