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.
Last week a group of Dutch scientists published a study providing further evidence of mindfulness’ ability to bolster creativity. Specifically they looked at if open awareness differed from focused attention in increasing divergent thinking
A common way of illustrating the idea behind statistical power in null hypothesis significance testing, is by plotting the sampling distributions of the null hypothesis and the alternative hypothesis. Typically, these illustrations highlight the regions that correspond to making a type II error, type I error and correctly rejecting the null hypothesis (i.e. the test’s power). In this post I will show how to create such “power plots” using both ggplot and R’s base graphics.
In this post I show some different examples of how to work with map projections and how to plot the maps using ggplot. Many maps that are using the default projection are shown in the longlat-format, which is far from optimal. Here I show how to use either the Robinson or Winkel Tripel projection.
In this post I show some R-examples on how to perform power analyses for mixed-design ANOVAs. The first example is analytical—and adapted from formulas used in G*Power (Faul et al., 2007), and the second example is a Monte Carlo simulation.
Can you tell when error bars based on 95 % CIs or standard errors correspond to a significant p-value? Don’t fret if you think it’s hard, a study from 2005 showed that researchers in psychogoly, behavior neuroscience and medicine had a hard time judging when error bars from two independent groups signified a significant difference
Why are physicists talking about 5-sigma, and what’s it got to do with statistics? In this short post I’ll explain what 5-sigma is and why it’s not a measure of how certain scientist are that they’ve found the Higgs boson