The model assumes independent errors. The data are repeated measurements on the same forty subjects. Nothing in the paper mentions this, the standard errors are consequently too small, and the p-value that carries the conclusion is a number produced by an assumption that was never true.

We edit what statisticians write — methodological papers and their theory, applied statistical analyses, simulation studies, model diagnostics and assumption checking, Bayesian and computational papers, statistical consulting reports, teaching material and textbooks, and theses, articles and grant applications. Our editors work on the assumptions and their reporting.

Every statistical procedure carries assumptions, and the failure is a paper that inherits them silently. An assumption unchecked is not an assumption avoided; it is a risk transferred to the reader. We work through these so the assumptions of each procedure are stated where the procedure is used rather than in a general methods paragraph, since a reader tracking one analysis should not have to assemble its conditions from elsewhere; so the check performed is described with what it showed, because "assumptions were checked" reports an activity and not a result; so the consequence of a violation is stated in the direction it matters, given that some violations inflate the type I error rate, some reduce power, and some do neither at the sample size in hand; so dependence in the data is addressed explicitly, as clustering, repeated measurement and spatial or temporal structure are the most consequential and most commonly ignored; so the response to a failed check is described as what was done, not as robustness in general terms; so a diagnostic plot is shown where it is informative rather than described; and so what the analysis assumes but cannot check is named. Papers written this way are the ones other people's analyses can be built on.

Everything you send is treated in confidence, including unpublished analyses, client data and consulting reports. We are editors rather than statisticians, and we offer no view on models, assumptions or results. What we can do is make the assumptions and their checks visible in the paper.