2
$\begingroup$

In JMP, I am building a regression model by using "Analyze"->"Fit Model" and choosing "Stepwise" for the personality. Once I click "Run" in the "Model Specifications" window, I get the "Fit Stepwise" window that allows me to specify how I would like my model to be built. So I hit "Go", and it builds my model.

My question comes in understanding the difference between what is displayed to me in "Current Estimates" and the values I get if I click on "Run Model" and look under "Parameter Estimates". It seems like these should be the same, but clearly JMP is doing something inbetween the Stepwise window and the Fit Model window.

Could someone please explain to me the difference between "Current Estimates" (in the Stepwise window) and "Parameter Estimates" (in the Fit Model window)?

$\endgroup$
  • $\begingroup$ I use SAS and occasionallly JMP. I don't know the answer for sure. So this comment is an educated guess. In the process of getting to the final model there can be several intermediate models considered with parameter estimates included. In the Stepwise window JMP is probably show parameter estimates computed at various steps in the procedure. $\endgroup$ – Michael R. Chernick Sep 13 '12 at 12:42
  • $\begingroup$ I had a similar thought. It looks like the Stepwise window does update the estimates those as you add/remove vars. JMP just doesn't have clear documentation on it, so I'm hoping someone can offer a definitive explanation! $\endgroup$ – cryptic_star Sep 13 '12 at 14:26
  • 2
    $\begingroup$ I don't know the answer to this, but I do know you shouldn't use stepwise to do variable selection. See my paper $\endgroup$ – Peter Flom Sep 30 '12 at 13:19
  • 1
    $\begingroup$ Agree with Peter Flom, stepwise regression should really be avoided. There are sensible alternatives, which are certainly available in SAS (e.g. LASSO, elastic net etc., e.g. see PROC GLMSELECT). I would hope that JMP would also point you into better directions than stepwise (really, you'd hope they would put a BIG WARNING MESSAGE like "WARNING: YOU SPECIFIED STEPWISE MODEL SELECTION, THIS METHOD IS WIDELY REGARDED AS INAPPROPRIATE" in the SAS log. $\endgroup$ – Björn May 7 '17 at 19:08
1
$\begingroup$

If any of the response or predictor factors in the stepwise regression have missing data for some rows the values of p-values, coefficient estimates, etc in the stepwise window can be incorrect. I'm not sure exactly what error is made in the JMP calculation but I have a dataset where I experiences this discrepancy, in it the stepwise calculations are based on my full dataset (n=72), but when I run the linear model on its own it is correctly based on only 50 samples and gives a different p-value, estimate, etc.

This has been a persistent element of JMP's stepwise calculations as I've experienced in from JMP 9 through 12. So it may be intentional to facilitate the comparison of predictors with different sample sizes.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.