Share your ideas for the JMP Scripting Unsession at Discovery Summit by September 17th. We hope to see you there!
Choose Language Hide Translation Bar
Highlighted
billi
Level IV

inverse prediction using quadratic model

 I calculated predicted x from linear fit using fit model > inverse prediction. Is it possible to do the same using the quadratic model. 

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted

Re: inverse prediction using quadratic model

I believe you are using the Bivariate platform. Use the Fit Least Squares platform instead. The Inverse Prediction command in this platform still does not work with quadratic models but there is another option. Follow these steps:

 

  1. Select Analyze > Fit Model.
  2. Select response column and click Y.
  3. Select factor column.
  4. Click Macros and select Polynomial to Degree.
  5. Change Emphasis to Minimal Report.
  6. Click Run.
  7. Click the red triangle at the top and select Factor Profiling > Profiler.
  8. Click the red triangle next to Prediction Profiler and select Optimization and Desirability > Desirability Functions.
  9. Control-click on the Desirability plot at the right end of the top row of plots.
  10. Change the goal to Match Target. Do NOT change the desirability values otherwise.
  11. Enter the target response value as the Middle value.
  12. Enter Low and High values that surround the target. These values are not critical.
  13. Click OK.
  14. Click the red triangle next to Prediction Profiler and select Optimization and Desirability > Maximize Desirability.

 

The inverse prediction appears as the red number below the profiler for the factor.

Learn it once, use it forever!

View solution in original post

5 REPLIES 5
Highlighted

Re: inverse prediction using quadratic model

Generally, no.

 

Capture.PNG

 

What is the inverse prediction for Y = 10?

Learn it once, use it forever!
Highlighted
billi
Level IV

Re: inverse prediction using quadratic model

@markbailey I am trying to calculate 'x' from 'y' using quadratic standard curves so I can compare with 'x' calculated using same standard data fir  to a linear model using jmp. I know I can do this graphically by using crosshairs tool  with overlaid linear and quadratic functions. Is there any other way to calculate 'x' from the quadratic response function using JMP to show the difference between this and corresponding values calculated using the linear model?

Highlighted

Re: inverse prediction using quadratic model

I believe you are using the Bivariate platform. Use the Fit Least Squares platform instead. The Inverse Prediction command in this platform still does not work with quadratic models but there is another option. Follow these steps:

 

  1. Select Analyze > Fit Model.
  2. Select response column and click Y.
  3. Select factor column.
  4. Click Macros and select Polynomial to Degree.
  5. Change Emphasis to Minimal Report.
  6. Click Run.
  7. Click the red triangle at the top and select Factor Profiling > Profiler.
  8. Click the red triangle next to Prediction Profiler and select Optimization and Desirability > Desirability Functions.
  9. Control-click on the Desirability plot at the right end of the top row of plots.
  10. Change the goal to Match Target. Do NOT change the desirability values otherwise.
  11. Enter the target response value as the Middle value.
  12. Enter Low and High values that surround the target. These values are not critical.
  13. Click OK.
  14. Click the red triangle next to Prediction Profiler and select Optimization and Desirability > Maximize Desirability.

 

The inverse prediction appears as the red number below the profiler for the factor.

Learn it once, use it forever!

View solution in original post

Highlighted
billi
Level IV

Re: inverse prediction using quadratic model

@markbailey Thank you for your response. It worked and I got results for the inverse prediction for linear model but I'll not be able to get inverse prediction for quadratic model in JMP. Correct?

Highlighted

Re: inverse prediction using quadratic model

Is what correct?

 

I followed the steps above using the Big Class data table. The weight is in the Y role and I added height and height^2 terms to the model. I defined the desirability function to match a target of Y = 120. Here is the result:

 

Capture.PNG

 

The inverse prediction is 66.3 inches.

Learn it once, use it forever!
Article Labels

    There are no labels assigned to this post.