I noticed this paper on astro ph today from Maxted et al. (2014) presenting a so-called Bayesian approach to inference of the mass and age of exoplanet host stars; I say “so-called” here because although there is huge emphasis in the Introduction on the advantages of priors in the Bayesian approach as compared to the maximum likelihood approach, when the authors present their analysis they use uniform priors almost exclusively such that the difference between the Bayesian and maximum likelihood methods is here reduced to the difference between the posterior mean and the maximum a posteriori (MAP) estimate. Tellingly, although their code is supposed to allow for non-uniform priors they quote the M-H acceptance-ratio as depending only on the likelihood ratio of the proposal and current position (within the prior support): i.e., embedding a uniform prior directly into their equations (it would be worthwhile for users of their code to make sure that this uniform prior isn’t so assumed in the accept-reject step of their source code). [Again this smells of poor refereeing by A&A.]

If I was refereeing the paper I would have asked for a proper comparison of the Bayesian and maximum likelihood approaches explicitly investigating the “temporal age bias” invoked in the Introduction, and I would have asked to see some investigation of a hierarchical model structure with shared population priors to be learnt with the data. ** But, then again no-one has ***ever* asked me to referee an astro-statistical paper …

### Like this:

Like Loading...

*Related*

“no-one has ever asked me to referee an astro-statistical paper …”

That’s a scandal. If I get one where I think you’d do a better job, would you mind if I suggested you as an alternative?

Of course, but I’m not sure that many astrostats papers fall in the coset of your expertise within mine. Somehow from astronomy journals I only get sent galaxy evolution papers which I increasingly often have to refuse, but I’ve already reviewed two malaria papers, having never published in the field!

This is asking for troubles… but I’ll be delighted to send some astrostats papers your way! 🙂

Bring it on 🙂