May 15, 2026
I asked Claude to summarize five years of my own peer reviews comments. Here are the kind of issues I usually flag.
You noticed that the p-values in a paper are close to .05. Does this make you a bad person? Can ‘circling p-values’ ever be a useful heuristic? My…
Let’s talk about errors in research, using myself as an example.
Creating studies that are powered to detect the smallest effect of interest… without collecting more data than you need to detect bigger effects
What happens when two researchers attempt to reproduce all the pre-registered studies with open data published in JCR?
How natural improvements in outcomes can lead us to form strong, yet incorrect, beliefs about how the world works.
If power analysis cannot be based on the expected effect size, what should it be based on?
Why power analysis, as traditionally performed, isn’t a good tool for choosing sample size.
Can p-values be close to .05 because researchers ran careful power analysis, and collected ‘just enough’ participants to detect their effect? In…
A short blog post showing the immense benefits of pre-registration on false-positive rates, even when the pre-registration is underspecified.
What can we learn from effect sizes, and under which conditions?
How can researchers respectfully and constructively flag inadequate statistical evidence when we they see it in papers? This post offers some…
A primer on evaluating statistical evidence, with a focus on p-curve analysis.
You have used the distBuilder library to collect data, now what? This post walks you through the basics of cleaning and analyzing distribution…
A short tutorial on adding “totals” to distBuilder, keeping track of how many balls are allocated in each bucket
A two-part blog post on outlier exclusion procedures, and their impact on false-positive rates.
A short case study showing how not to deal with your outliers, featuring a recent paper published in psychology.
Do you wonder how respondents are using distBuilder to construct their final distribution? This blog post shows how you can record the history of…
You want to test how well people can learn a normal distribution. How can you make sure that the discrete number of values that you will show to…