Estimate Cohen's d* after mixed models

[Sunniva]

I have computed a mixed model ( see the code below) to estimate treatment effects in a RCT study. I have three measure points (time 0, time 1 and time 2). I would like to get the effect size (Cohen's d) for time 1 and time 2. But I can not figure out how I can estimate Cohen’s d after a mixed model. I am familiar with the esize command, but using esize I can not take into account the dependency in my data ( three level model).
I would really appreciate if someone could tell me how to estimate Cohen’s d after mixed model.
Many thanks.


mixed dependent i.time 1.tretament#i(1/2).time ||classid:||id:

esize twosample dependent if time ==1, by(treatment)

esize twosample dependent if time ==2, by(treatment)

[Joseph Coveney]

Is Cohen's d even defined in the context of such a mixed-effects regression model? If so, then you could show its definition in this context, and someone might be able to help you compute it from the parameter estimates of your fitted model.

[Sunniva]

Dear Joseph, this is a good question, unfortunately I don't really know the answer. I have to figure this out. If someone have estimated chohens d after mixed model , please let my know.

[Joseph Coveney]

My recommendation is to skip Cohen's d—recourse to which seems like sheer desperation—and go for a substantive interpretation of the magnitude of the estimated effects and their precision, the latter from the standard errors of the estimates given just to the right of the parameter estimates, themselves, in the regression table. Interpret the magnitude of the parameter estimates by reference to your subject matter knowledge and not as transformed into some value whose yardstick is a semi-arbitrary rule of thumb.

[Sunniva]

That’s probably the best solution.