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
- /
- General Programming
- /
- How to run conditional multinominal logistic regre...

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

04-07-2015 12:07 PM

Hi all,

I have a question about multinominal logistic regression, if possible? My outcome has 3 unordered levels, so I want to use polytomous (multinominal) logistic regression. But my patients are matched by age, so I need to run a conditional polytomous logistic regressoin. However, SAS or STATA seem not to support the conditional polytomous logistic regression.

For example, my SAS codes are:

**Proc logistic data=w descending;Class drug/param=ref ref=first;Model outcome= drug /link=glogit;Strata age;Run;**

Then SAS cannot proceed:

**ERROR: The LINK=GLOGIT option is not allowed when a STRATA statement is specified.**

I have been searching online for long time, but no gain. The discrete choice model (using proc phreg) does not fit my data, however. My independent variables are all patients' individual characteristics..

Many thanks for any help or suggestion!!

Mileo

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

Posted in reply to mileo

04-07-2015 12:36 PM

I know we can use PROC PHREG for conditional logistic regression, but I'm not sure it expands to conditional multinomial

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

Posted in reply to Reeza

04-07-2015 01:31 PM

Thanks so much, Reeza!!

I follow the example you gave:

**Proc phreg data=w descending;Model outcome*outcome(0)= drug / ties=discrete;Strata age;Run;**

However I don't think it is appropriate for multinominal logistic regression, as the results only show the effect of drug on a binary outcome (with df=1). Typically the multinominal analysis should show different effect on different level of outcomes (in my case, df=2)..

Anyway, thanks so much!!

mileo

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

Posted in reply to mileo

04-07-2015 01:48 PM