What is it about?

You can already test whether or not two distributions are identical with tests like Kolmogorov-Smirnov, Cramer-von Mises, and Anderson-Darling. For example, you can test if control group and treatment group distributions differ (i.e., did the treatment have any effect?), or two socioeconomic group distributions, etc. But sometimes you also want to know *where* the distributions differ. The new distcomp command reports ranges of values where two distributions are statistically different, while controlling the false positive rate appropriately. It's based on my 2018 Journal of Econometrics paper with Matt Goldman.

Featured Image

Why is it important?

The distcomp command lets you learn not only if two distributions differ, but where they differ.

Perspectives

I implemented distcomp in Stata because it's one of my favorite new statistical/econometric methods that I've developed.

David M Kaplan
University of Missouri

Read the Original

This page is a summary of: distcomp: Comparing distributions, The Stata Journal Promoting communications on statistics and Stata, December 2019, SAGE Publications, DOI: 10.1177/1536867x19893626.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page