868412_en-US

liliconcom

Lack of Fit in Fit Model

Okay - interesting question I wanted to pose since we don't get Lack of Fit questions that often (at least I don't). 

A customer asked me about why a Lack of Fit would or wouldn't show up. 

The model (when parameter 3 is included) includes LOF reports:

But , when that Parameter 3 is taken out the LOF report disappears:

 

Looking at a scatter plot matrix, it looks like the model has the needed replicates to do a LOF test, so we're not sure what criteria JMP thinks we're missing.  Thoughts?

 

 

Re: Lack of Fit in Fit Model

Lack-of-fit requires TWO conditions: pure replication and more conditions than what is necessary for estimating your model. When you remove parameter 3, you are left with three levels for parameter 1. You need three levels to estimate that quadratic, so a lack-of-fit is not possible.

Re: Lack of Fit in Fit Model

Also credit to @Di_Michelson  who talked me through the same thing Dan said while we couldn't comment yesterday.