turn on suggestions

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

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- SAS Procedures
- /
- obtain performance metrics for proc plm

Topic Options

- 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
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

01-04-2017 04:53 AM

This code:

proc plm restore=SomeLib.ModelParams; score data=SomeLib.TestData out=SomeLib.Predictions; run;

produces the dataset SomeLib.Predictions which contains the predictions of the model against the dataset SomeLib.TestData.

Just wondering, can I use proc plm to obtain performance metrics such as:

Root MSE

Dependent Mean

R-Square

Adj R-Sq

AIC

AICC

SBC

ASE

Accepted Solutions

Solution

01-05-2017
07:36 AM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to csetzkorn

01-04-2017 08:57 AM - edited 01-04-2017 09:10 AM

Not really. Those statistics have to do with evaluating the suitability of the model fit, so they come out of the regression procedures. They don't change after the model is fit. In contrast, PROC PLM is intended for for post-fitting analyses (after you have settled on a model) such as scoring new data or testing for treatment differences.

PROC PLM does provide some model statistics if you use the SHOW statement. For example, you can submit SHOW FITSTATS, which provides the MSE and ErrorDF statistics. So you can get figure out the RMSE. But I don't see an easy way to get AIC, BIC, etc

All Replies

Solution

01-05-2017
07:36 AM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to csetzkorn

01-04-2017 08:57 AM - edited 01-04-2017 09:10 AM

Not really. Those statistics have to do with evaluating the suitability of the model fit, so they come out of the regression procedures. They don't change after the model is fit. In contrast, PROC PLM is intended for for post-fitting analyses (after you have settled on a model) such as scoring new data or testing for treatment differences.

PROC PLM does provide some model statistics if you use the SHOW statement. For example, you can submit SHOW FITSTATS, which provides the MSE and ErrorDF statistics. So you can get figure out the RMSE. But I don't see an easy way to get AIC, BIC, etc

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Rick_SAS

01-04-2017 12:21 PM

Thanks but they change when I apply the model to new data (e.g. to assess when the model start to diverge - e.g. adjusted R squared decreases). I will take a look at SHOW.