mean command postestimation for variance

[Sun Lee]

Hello,

I have a question regarding accessing variance estimates from mean command. In my data, there are treatment and control groups over 6 years. I want to test if the variance estimates are different between two groups in each year.

mean outcome, over(year group)
estat sd, variance

After this very simple code, I get mean and variance estimates for two groups separately in each year. I am trying to use test command to do a Wald test to see if the variance estimates of the two groups are different. I acknowledge this is just an approximation, and other tests may be better, but this is irrelevant for my question.

When I use a regression, I used something like

test [RP1]var(treatment group) = [RP2]var(control group)

to do the Wald test. How can I do this after mean command?

Thanks,

Sun

[Clyde Schechter]

There is no Wald test here because the output from -estat sd- does not include anything like a standard error of the group variances.

If you want to do a test of equality of variances in the groups, you can do an F-test. (V1/(n1-1))/(V2/(n2-1)) has an F distribution with n1, n2 degrees of freedom under the usual normal theory assumptions. V1 and V2 can be pulled from r(variance) and the n's can be pulled from e(_N) after -estat sd, variance-.