How does JMP handle weights on single data points in regression?

Question

I wish to do a weighted regression of 16 data points, some of which are means of several measurements, but some are single measurements. Normally, I'd weight by 1/variance. But what does JMP do with weighting the single (unreplicated) data points? Do they get no weight, maximum weight, or something in between (like 1/total n weight).

Thanks folks,

Steve

Phil_Kay · Answer

Hi Steve,

Your question is not clear to me.

"what does JMP do with weighting the single (unreplicated) data points?"

Do you mean what happens when you use 1/variance for the weighting? Because, of course, with a single data point you have no estimate of variance. So what would you be using for 1/variance in that case? 0? Using 0 weighting would mean that the data point is excluded from the analysis.

Let me know more about what you were thinking. Perhaps you could provide example data. Also let us know what platform in JMP you are using.

Regards,

Phil