Why `naive_se`?
Closed this issue · 0 comments
ParadaCarleton commented
It seems a bit unusual to display naive_se
by default, given it's strictly inferior to the MCSE measure. Mostly I'm concerned about people misunderstanding it. There's lots of things that naive_se
could plausibly be confused with -- the standard error of the sample mean for the data, the standard deviation of the posterior distribution, or it might just be taken to be a good estimate of the MCSE, which it often isn't.
Another problem with naive_se
came up in the course of adding support for weighted means, which are required for nested sampling. It's not at all clear to me what the "Naive" error here would be, since there's no autocorrelation.
Would it make sense to remove it?