Odds ratio interpretation
Message/Author
 Eoin O'Connell posted on Tuesday, August 25, 2009 - 9:01 pm
LOGISTIC REGRESSION ODDS RATIO RESULTS

U7 ON
U2 1.109
U3 0.896
U4 0.913
U5 1.227
U6 2.669
U9 0.998
Y2 0.422
Y3 0.873

When I run a logistic regression, I obtain the following ORs. However some of the categories I used have 3 levels (u2, u3 and u4) but just one ratio. How can I calculate the odds for each of the 2 other categories based on the first? Should I be using the thresholds in the model results?

Also, how can I get Mplus to do two separate regressions, split by gender? Is it by using the CLASS function?

Thank you,
Eoin
 Linda K. Muthen posted on Wednesday, August 26, 2009 - 12:05 pm
There should be two odds ratios for a variable with three categories. Please send the full output and your license number to support@statmodel.com.

You can use the USEOBSERVATIONS option to do an analysis for a subset of the sample. You can use the GROUPING option to do multiple group analysis.
 Dallas posted on Thursday, April 14, 2011 - 4:58 am
Linda,

Good morning. In the MLM setting, I easily understand how to compute the odds ratios from the fixed effects. In addition, I realize I can create a CI by first computing the CI for the beta coefficients and then exponentiating them. But, what formula would you suggest for estimating the ORs SE?

Thanks.
 Bengt O. Muthen posted on Thursday, April 14, 2011 - 8:06 am
If you use Model Constraint to express the odds ratios from the parameter estimates you get SEs automatically. This uses the Delta method.
 Dallas posted on Thursday, April 14, 2011 - 11:12 am
Thank you! I haven't used that option before. Could you direct me to an example? Thanks.
 Linda K. Muthen posted on Thursday, April 14, 2011 - 11:51 am
See Example 5.20 and MODEL CONSTRAINT in the user's guide.