Keith Goldfeld, DrPH, MS, MPA, member of the executive committee of the IMPACT Design & Statistics Core, recently published a blog post inspired by discussions with collborators from the IMPACT Collaboratory.
Goldfeld discussed how the question of variable cluster sizes has come up a number of times in recent discussion with IMPACT Collaborators about setting the sample sizes for proposed cluster randomized trials, Goldfeld explains that when working with a fixed overall sample size, it is generally better (in terms of statistical power) if the sample is equally distributed across the different clusters. Highly variable cluster sizes increase the standard errors of effect size estimates and reduce the ability to determine if an intervention or treatment is effective.
Goldfeld realized that there is no easy way to generate the desired variable cluster sizes while holding the total sample size constant using simstudy, his preferred simulation package. In response to this, he developed a simple solution that is available for download on the blog.