@amoeba It's a common expression in the UK and to some extent in other places.

I do use it myself sometimes

@Kodiologist I'm not aware of either any policy against that specific problem (other than "please don't cross-post") -- nor a built-in mechanism for closing cross-site duplicates (outside a typed-in closure reason)

I'm certain someone has asked a question like this before but I can't find a good duplicate. stats.stackexchange.com/questions/215560/…

@GeneralAbrial: Me too. Was it something to do with post-codes?

No, that was categorical predictors with many levels.

This is a bit of a vague question, but I'm not quite sure what I'm after. Let's say I have some 1D data that represents physical measurements that might fall into one or two groups. I know it's one or two groups, but for any given set of measurements, I don't know which. Is there a way to parameterise how likely it is to be one or two groups, so I can make a decision based on that likelihood?

I've been looking at eg. Jenks natural break optimisation, but as far as I can tell, that requires you to know how many groups you want and optimise the boundaries from that.