It seems astronomers just can’t get enough of the K-S test: in this new paper by Klesman and Sarajedini the authors test for differences between the radial distributions of AGN and normal (non-AGN) galaxies in clusters using the K-S test and drive their ship straight into the iceberg known as “multiple comparisons“. For each of 14 clusters they perform the K-S test on each of three nested subsets of the data constructed through increasingly inclusive sample selection criteria. Four of the 14 clusters have a <0.05 p-value in at least one of the three subsets which they take to be significant; but no subset has more than two clusters with <0.05 p-values, which we would expect >10% of the time under the null.

### Like this:

Like Loading...

*Related*

I bet it’s because the KS test is in Numerical Recipes.