Hi All, I just wrote this blog post adapting methods taught by @drob about using the Beta Distribution to estimate probability. Would love feedback in general and on any of the following:
- If you are not already well familiar with Bayesian methods, does the rationale for the workflow make sense?
- If you are familiar with Bayesian methods, does the selection of the prior seem reasonable? What about rescaling the alpha and beta parameters to make the prior less informative?
- Does the last figure and its "tug-of-war" analogy make sense?
- Does it seem reasonable to use pop_under_five as a proxy for the beta parameter since we don't know the true denominator of live births (and hence survivals)?
I'm happy to exchange feedback for something you're working on too.