This will be great to address the typical comments by reviewer “you should justify why you fitted a Bayesian model” (they never once justified why they fitted a frequentist model)
ooh, is this a frequentist vs Bayesian kind of thing? I've never actually understood the distinction, other than that my college courses on probability say that these 2 views exist and apparently cause disagreements...
You mean CLT? You don't need Bayesianism to deal with wonky data generating processes with weird mass points. Plenty of frequentist methods for that stuff too. Though I guess Bayesianism is probably a bit easier in those cases.
If I were to distill my critique of Bayesianism down to a sentence or two: It's a lot more work (both computational and otherwise), often misused, and to the degree the result differs from what you get with frequentist statistics, the difference is all about the prior you specified.
I think my main criticisms are: - p-values are poorly understood (and unsatisfying). - people are not doing power analyses before starting experiments. Which is more of a people misusing Frequentist statistics versus an inherent flaw, but it is a persistent problem.
The stats training in my political science department is all frequentist garbage Because "that's what the journals look for," you see
A nice (simplified) video on the differences between frequentist and Bayesian approaches to probability and statistical inference. #stats (What it explains quite well is that while you can bias you analysis in Bayesian, people will find out easily. More easily than in frequentist analyses...)
YouTube video by Very Normal
It’s sort of like if you took a random sample of papers that used Bayesian statistics and compared them to papers using frequentist statistics. The Bayesian papers would look more sophisticated, but mostly it’s because on average, people who know Bayesian stats usually have more stats training.
Haha yeah, haven’t seen much of the ol’ frequentist vs. bayes stuff. Just give it time!