How to calculate odd ratios and predicted probabilities from a multi-level logit model?

[Olivier Mordant]

Hi all,

I have fitted a multi-level logit model exploring the risk for infection (binary outcome) across communities (clusters).

xtmelogit infection || community:, variance

My queries are as follows:
1) I have fitted the model and adjusted for sex and ethnicity. The model shows me their coefficients but not odd ratios - how do I convert them into odd ratios for each of the categories of the variables (e.g. different ethnicities in the ethnicity variable)?

xtmelogit infection sex ethnicity || community:, variance

infection | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
         sex |  -.2615564   .2184337    -1.20   0.231    -.6896787    .1665658
   ethnicity |  -.0538283   .0143838    -3.74   0.000    -.0820201   -.0256365

2) I would like to calculate the predicted probability for infection for residing in each of the community (i.e. at the cluster level). I have use the following command, but I am not sure how to proceed therefrom in terms of calculating 95% CI and creating a caterpillar plot. Any advice?

predict pr

Thanks so much!
Oli

[Rich Goldstein]

first, -xtmelogit- was replace by -meqrlogit- many versions ago; as the FAQ explains and requests, you need to tell us what version you are using if it is not the current version (18)

second, each of these commands has an option "or" that will do what you want as explained in the help file for each