turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Does anyone know how p value of Actual by Predicte...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 7, 2014 12:38 PM
(1559 views)

Hi

After running a linear regression using the Fit Model platform, JMP displays P value (as P), RSq and RMSE on the bottom of the a Actual by Predicted Plot. I would like to understand how P value is calculated and what it means.

I ask this because in a simple Y =aX+b case, the p value is identical to the p value of the X **but** if there are more Xs (e.g. Y = aX1+ bX2 + c) then, p value is something else. I would like to know the meaning and calculation of p value of Actual by Predicted Plot in multiple regression situations. Any comments would be very welcomed.

Thank you.

W Kim

3 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 7, 2014 8:29 PM
(1183 views)

W Kim,

The p-value underneath the Observed vs Predicted plot is the p-value from the overall analysis of variance report. This report is the overall test to determine the significance of the entire model. For example, if the model is Y=B0 + B1*X1 + B2*X2, the null hypothesis for the overall ANOVA would be Ho: B1=B2=0. So this is testing if the proposed model does better at explaining the variance in the response than the overall mean.

The calculation of the p-value is more difficult to describe in this brief response, but it is the probability of seeing the associated observed F-statistic under the null hypothesis. You calculate the p-value by finding the area under the F-distribution with the proper degrees of freedom to the right of your observed F-statistic. Thus, if the p-value is low, there is a small chance of observing the data we have. That implies that the null hypothesis is unlikely to be true.

You may wish to look into some regression books for more specific details.

I hope this helps.

Dan

Dan Obermiller

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Aug 21, 2014 6:30 AM
(1183 views)

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Aug 21, 2014 6:50 AM
(1183 views)

Also you might want to look at the JMP training course "Anova and Regression" which discusses exactly these types of issues.

Dave

-Dave