[tpoacher]

The "small sample size" has been drilled to most people as a knee jerk red flag indicating a statistically unreliable conclusion, but this is not always the case. Depending on the desired outcome, a small sample size may actually increase the reliability of the result. This seems like one of those cases.

Think of it like this. The aim is to show a statistically significant difference. With a large enough sample, you will eventually show significance (hence Fisher's famous anecdote of "get more data"), but the effect size may be trivial. Whereas, with a small sample, only non-trivial effect sizes will achieve significance, and therefore achieving significance with such a small sample tells you something about the nontrivial extent of the effect size.

So you could argue that the extent of methylation change was so large, that it achieved significance in a sample even as small as n=13!