So I arXived my brief research note on a generalization of the Savage-Dickey Density Ratio (SDDR) to general probability spaces (including those not admitting a Lebesgue reference measure) yesterday (so it’ll be out in today’s mailing later). But this morning I received the Editor’s decision: rejected without even sending to referees. The reasoning was thus:
“The paper builds on the work of Marin and Robert and presents a generalized Savage-Dickey Ratio. The paper is well written and its arguments easy to follow.
However, its contribution is rather increments. I believe it is better suited for a journal focusing on short communications.”
Which is all a bit confusing for me because the description of EJS at the top of its website is:
“The Electronic Journal of Statistics (EJS) publishes research articles and short notes on theoretical, computational and applied statistics.”
As for the slightly pejorative “rather increments”(sic) comment: fair enough I guess, though the Marin & Robert (2010) paper from EJS (which is why I submitted there) could have been described as incremental, and it’s my impression that a lot of good scientific papers are. Also, it’s my feeling that Bayesian model selection in non-Lebesgue density spaces (e.g. Dirichlet process priors, Gaussian process priors), though little studied to-date, is likely to become quite a hot topic in the next few years.
Oh well, at least it’s on the arXiv (in a couple of hours), and should make a quick read for anyone interested in Bayesian model selection: if you can read it before finishing your morning doughnut, then you can treat yourself to another! 🙂