Hi,

I am designing a RCTs with 36 clusters. One of my colleague criticized that my study is severely under-powered and urge me to change the design. It was confirmed when I computed power with power or sampsi command in Stata.

I have found a paper from F.Burlig et al (2020), in which they introduce a new Stata module for power analysis called pcpanel. I have tried pc_simulate command and the results are too good to be true.

It turns out that I am able to run RCTs with 36 clusters. In the text, they basically said that current power approach for short panel is too conservative, and we are recruiting more units than we are actually needed!

However, actually I am doubt with my interpretation. Does anyone here have read this paper and have the same interpretation? Is it true that existing power command produce too conservative outcomes?

Thank you so much for reading the question.

Regards.

Citation:

Burlig, Fiona, Louis Preonas, and Matt Woerman. 2020. Panel data and experimental design. Journal of Development Economics 144:102458.