Medcouple statistic

Mark_Zwald

A customer recently asked if JMP supports the medcouple statistic (https://en.wikipedia.org/wiki/Medcouple).

It appears the answer is no.

Do we have a best practice for customers to follow to identify outliers taking into account skewed distributions? 

Thanks,

Mark

Byron_JMP

a pretty good method is to fit all continuous distributions (probably want to use the legacy filters, those are a lot easier to work with) then save the transformed values. Check the box plots on the transformed values and see what you think.   The Johnson and SHASH distributions are fairly flexible and tend to fit the tails (and maybe over fit?) well.