All,
this is just a watchout if you use the distribution platform on categorical data with some observations hidden/excluded from the analysis in JMP 13 (and also V12). I reported this to the support team and they confirmed working on a hot fix for this, therefore implicitly confirming this is a bug I guess. No need from my end to respond, I identified workarounds (see below) that worked for me, so this is just to raise awareness as it might not always be immediately apparent from an analysis.
So I wanted to drop observations from a simple analysis on nominal data. Yet, the excluded row is reported as a missing value in the analysis, so the “Total” + “N Missing” is 7 rather than 6 as it should be. For that, it doesn’t matter whether the excluded observation is anyhow missing, or whether it has a value.
The issue occurs only for nominal or ordinal data, not for continuous one, and is reproducible on my machine as well as that of several JMP representatives. A simple data set with script is attached, showing the issue.
Possible remedies (there are certainly others):
- delete the observations entirely from the data
- use a WHERE statement (in JSL) or a BY variable - the latter might not always be possible directly, so you may have to create a variable that specifies which variables you want to drop.
Cheers, Michael
