Programming the statistical procedures from SAS

Multinomial logistic regression: reference group

Reply
Frequent Contributor
Posts: 131

Multinomial logistic regression: reference group

Hello sir,

I used PROC LOGISTIC to run a multinomial logistic regression for a response variable with 3 levels. The result for an explanatory variable 'region' is as below:

                        Analysis of Maximum Likelihood Estimates

                                                              Standard          Wald
Parameter             group    DF    Estimate       Error    Chi-Square    Pr > ChiSq

region        01          1         1      0.6996      0.2886        5.8762           0.0153
region        01          2         1      0.9066      0.0212     1828.5791        <.0001
region        02          1         1      0.3783      0.2958        1.6362           0.2009
region        02          2         1      0.8891      0.0214     1733.4656        <.0001
region        03          1         0      0.7229           .               .                   .
region        03          2         0           0              .               .                   .

The result for region 03 looks unsual to me. The region=03 acts as the reference group. So df=0. For group=2, there is no estimate which is correct. But for group=1, the estimate actuall has a value, not 0. This is problematic given the meaning of reference group.

Is there any reason why this happend? Thanks in advance for your instruction.

Ask a Question
Discussion stats
  • 0 replies
  • 138 views
  • 0 likes
  • 1 in conversation