The ELS blog has had a thing about interaction terms in regression-y models for a while now (see e.g. here and here and here). For readers outside the quantitative-political-science ecosystem, there has also been an ongoing set of... shall we say... discussions in the field about specifying and interpreting interaction terms in generalized linear models for twenty years or so. A recent paper by Tsung-han Tsai and Jeff Gill is the latest weigh-in on the matter, and also manages to be an excellent, clear exposition of the use and interpretation of such interactions. Since the journal in which it appears is relatively new, I thought I'd bring it to people's attention.