Calculating minimum detectable difference for cluster RCT

[Lily B]

Hello Stata Forum,

I’m trying to calculate the minimum detectable difference for a number of continuous and dichotomous outcomes. I’d like to get a few estimates for different levels of power and alpha. I have a cluster RCT with 25 clusters in the control group and 30 clusters in the treatment group. The size of the clusters is also different by treatment and control group.

I’ve successfully used clustersampsi for other cluster trials before, but it does not work in this case as it does not seem to support a design with different number of clusters per group. I also tried clsampsi, but this command only reports estimated power and not minimum detectable difference.

Any other commands out there to calculate detectable difference for cluster trials? I’ve searched everything on the Stata Forum, Stata Archive, Stata Help and Google and can’t find a command that can output what I need.

Thanks,

Lily

[sabine.gabrysch]

Hi Lily and others,
Did anyone find a solution to this yet? I have the same issue.
I guess I can use Optimal Design software, but would be nice to be able to do this in Stata...
Thanks,
Sabine

[Carole J. Wilson]

It seems that clustersampsi (Stata Journal) is able to calculate the detectable difference in cases where the cluster sizes vary, but not when the number of clusters differs across arms. That is, you can specify the average cluster size and the coefficient of variation in cluster size. Specifics can be found in the Stata Journal (13(1): 114-135) article written by the program's authors: http://www.stata-journal.com/sjpdf.h...iclenum=st0286

For example:

clustersampsi, binomial detectabledifference p1(0.3) m(20) k(25) rho(0.05) size_cv(0.9)