BookmarkSubscribe
Choose Language Hide Translation Bar
Highlighted
dhawker
Community Trekker

Adding formula to ternary plot gives me 'Mixture sum is not consistent' alert

I have a mixture design experiment with results as pass or fail (0 = pass, 1 = fail). I fit a nominal logistic model for the results, and get a pretty good fit that I would like to plot on a ternary diagram of the mixture space. I've tried using the probably formula columns from the model as the contour formula when setting up the plot, but I keep getting a "Mixture sum is not consistent" error and no contour (just a plot of the different mixtures). I can't find any information on what this alert means so that I can hopefully address it, does anyone have any experience with this error? Any suggestions on how to plot my pass/fail prediction as a contour on a ternary diagram? Thanks!

0 Kudos
1 ACCEPTED SOLUTION

Accepted Solutions

Re: Adding formula to ternary plot gives me 'Mixture sum is not consistent' alert

To diagnose the problem, if you create a new formula column with the sum of the mixture components, does the result always equal 1?

 

(A quick way to perform this check is to first select all the data columns with the mixture components, right-click on the selection, and select New Formula Column > Combine > Sum.

Learn it once, use it forever!
2 REPLIES 2

Re: Adding formula to ternary plot gives me 'Mixture sum is not consistent' alert

To diagnose the problem, if you create a new formula column with the sum of the mixture components, does the result always equal 1?

 

(A quick way to perform this check is to first select all the data columns with the mixture components, right-click on the selection, and select New Formula Column > Combine > Sum.

Learn it once, use it forever!
dhawker
Community Trekker

Re: Adding formula to ternary plot gives me 'Mixture sum is not consistent' alert

This worked! Some of my sums (generated by JMP) were 1.0000001 so I just adjusted and it let me get past the error. Now to figure out a way to plot a contour based on Prob[0] and not Lin[0] so I can visualize my results...
0 Kudos