The next version of powerlmm (0.4.0) will soon be released, besides bug fixes this version also includes several new simulation features. In this post I will show two examples that cover the major new features.
Non-randomized comparisons are common in RCTs. In this post I show some examples of confounding and collider bias, using treatment adherence as an example. I present a small simulation study that show that common regression models used in clinical psychology, makes little sense, and that Bayesian instrumental variable regression can be easily fit using the R package brms.
Over the summer I've been working on finishing my new R package 'powerlmm', which is now almost complete. It provides flexible power calculations for typical two- and three-level longitudinal linear mixed models, with unbalanced treatment groups and cluster sizes, as well as with missing data and random slopes at both the subject and cluster-level.
In this post I will use the theoretical and empirical sampling distribution of Cohen’s d to show the expected overestimation due to selective publishing. I will look at the overestimation for various sample sizes when the population effect is 0, 0.2, 0.5 and 0.8. The conclusion is that you should be weary of effect sizes from small samples, and that the issue is rather with type M (magnitude) errors than type I errors. At least is clinical psychology the pervasive problem is overestimation of effects and not falsely rejecting null hypothesis.