5 ms·
The nice thing about Bayesian analysis is that you have to make your priors explicit. That’s what this is, a prior. It doesn’t have to have a derivation; it can
by GlenTheMachine 6y ago
The nice thing about Bayesian analysis is that you have to make your priors explicit. That’s what this is, a prior. It doesn’t have to have a derivation; it can just be a guess. The important thing is that you’re writing your guess down so everyone can see it.
You can then also do a sensitivity analysis to figure out how much your conclusions change if you modify your priors. So if you, the reader, think the priors are wrong, then you can change them and re-do the analysis.
I think the most interesting thing about this analysis is exactly that: we can look at the priors and come up with a principled conclusion. We can then argue about whether the priors are right.
- casion 6y agoThe major issue here is that the sensitization of priors can be a marketing ploy. It's very easy to "make guesses" that present your conclusion, throw in a paltry amount of money, then make bank off the publicity. This possibility poisons the well for the entire process. If they can make money off this even if they are completely off-base, then it's not rational to build up the trust necessary in their process to engage with their model.
- GlenTheMachine 6y agoIt’s not clear to me how they make money other than by winning their bets. What am I missing?
- casion 6y agoThis is a company that is selling a product, and they're receiving a GREAT deal of publicity about their utilization of a methodology core to their offerings.
- GlenTheMachine 6y agoWhat’s the product?