"Multiverse Analysis" is a terrible name, that can only introduce confusion, even if it increases readership. This methodology is more about machine learning th

"Multiverse Analysis" is a terrible name, that can only introduce confusion, even if it increases readership. This methodology is more about machine learning than it is about covering all possible worlds. I'm also suspicious of the entire approach. In order to compare two different methods of analysis, you have to do some kind of mapping or weighting or meta-analysis that tells you why comparing two numbers makes any sense at all. This result must be tautological, in the sense that every expected value and variance can be computed and compared to another of the thousands of methods employed, and when we pour in the data it has to fall within the expected ranges, otherwise we know we've made a mistake in the original calculation.