Ordered logistic regression PreviousNext
Mplus Discussion > Categorical Data Modeling >
 Anne de Bruijn posted on Thursday, July 28, 2016 - 1:22 am

I'm quite new to Mplus and I'm trying to conduct an ordered logistic regression on an outcome variable with four groups. I did the same in SPSS before and there I got results in which a reference category was compared to all other categories (i.e. whether an independent variable predicts categorization into group 1 vs. reference group, group 2 vs reference group and group 3 vs reference group). I was expecting the same results in Mplus, but as far as I get it, results show values associated with a >50% of transition from one level to a higher level. I'm actually interested in whether my IVs predict categorization in the reference group or one of the other groups. Is this possible in Mplus? Or am I completely off in my interpretation of the results. I tried using dummy variables in which two groups were compared, but I'm not sure on how to do this as well..

Thanks in advance!
 Bengt O. Muthen posted on Thursday, July 28, 2016 - 9:33 am
When you say "an ordered logistic regression on an outcome variable with four groups", I assume you mean that the outcome variable of the logistic regression has four ordered categories.

The comparisons you mention sound more like unordered (nominal) logistic regression. The interpretation of ordered logistic regression is described in our new book:

 Anne de Bruijn posted on Thursday, July 28, 2016 - 11:58 pm
Hi Bengt,

Thanks for your response. I indeed have an outcome variable with four ordered categories. I was also noticed that the way I want to interpret my results is more in line with nominal logistic regression, but I'm not sure whether that's the best thing to do as my data is ordered. Do you have any suggestions?
Thanks again!
 Bengt O. Muthen posted on Friday, July 29, 2016 - 12:20 pm
For ordinal outcomes you look at the probability of the highest category versus remaining categories, the 2 highest categories versus remaining categories, etc. Those have the same log odds ratios. See the book for a worked example.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message