Whenever I see campaigns such as this, I'm reminded of the 2021 tweet from Surrey Poilce

Plausibly, this could affect animals more strongly (less domesticates overall, larger areas for the same population)

I've seen an argument made that Americas are "tall" so you get only a narrow segment of each climate zone -> domesticates spread less widely -> less opportunity for further breeding. Contrast with Eurasia where domesticates did easily spread across large areas allowing for more innovations.

I also should point out that in the example I've shown, the true parameter values are perfectly compatible with both priors and the posteriors don't change in any meaningful way when the prior changes, so BFs are uniquely sensitive here.

I totally agree. What I am saying is that you cannot generally handwave away the choice of prior for _nusiance parameters_ when computing BFs as is often done. Prior for the parameter of interest is already hard. Good prior for the nuisance parameters IMHO substantially higher bar.

No, I have two sets of two models, in both H0: intercept only, H1: intercept + one random intercept. The two sets differ only in prior for intercept and residual sd (default vs. weakly informative). Identical dataset. BF H1-default/H0-default is 4.23, while BF H1-weak/H0-weak is 2.83.

While the same change in prior makes no visible difference to the posterior.

I think the received wisdom on priors for nuisance parameters deserves some pushback. E.g., when testing for presence of random effects, moving from default improper prior on intercept + residual sigma to a weakly informative prior can change the BF by 50%. gist.github.com/martinmodrak...

This makes direct BF interpretation tricky. Their use needs can still be defended as heuristics/game or buttressed with a ton of prior sensitivity analyses, but that is IMHO a harder sell 2/2

Agree that the software is there. It is still "hard" to provide meaningful priors, _including parameters that are not of direct interest_ (despite popular belief) and IMHO you often cannot do that (e.g. you typically have beliefs about joint behavior of all parameters, but an independent prior) 1/2

I think this is a bit complicated by the languages that extend some kin terms outside of kinship, but don't extend the implied social obligations to match (e.g. aunt/uncle in some Slavic languages)...

Fun fact: in Slovak, you use the word for aunt/uncle also as an informal variant of lady/sir. As in "the aunt driving the tram" or "some uncles were arguing in the shop". So a _very_ inclusive term. In Czech it is less extreme, but the words extend to friends of your parents.

"It is not likely that there will ever be a point of time at which all fundamental debates in all fields of science reach a nice closure. Moreover it is not even clear that pluralistic phases of sciences are confused and uncertain, and therefore inferior, to the more unified phases. >>