Reference category in mi estimate logistic

[Anna Seijmonsbergen]

Dear all,

I would like to do a logistic regression with multiple imputed data. My syntax is:

mi estimate, or: logistic csection_02 i.country parity_01 i.mat_age_01 i.ethnic_01
contrast g.country, or level(99)

The first line is running well, but the 'contrast' part is not working. The 'contrast' part works in logistic regressions with non-imputed data. However, I think my syntax is not correct for mi data. I would like to change the reference category of the variable 'country' in the weighted mean of csection of the countries.
Example: the weighted mean incidence of a caesarean section between the 13 included countries is 24%. This 24% should be the reference (OR of 1.0). How can I do this?

Adding b(...) to the first line is not working.

Many thanks in advance!

Best wishes,
Anna

[daniel klein]

I have not understood your verbal description, but you might find mimrgns (SSC) helpful.

mi estimate : logistic ...
mimrgns g.country , eform

Note that both options for logistic are unnecessary because (i) logistic always displays odds ratios and (ii) mi estimate has its own way of displaying output, thus reporting options (including level(99)) for logistic are ignored.

Best
Daniel

[Anna Seijmonsbergen]

Dear Daniel,

Many thanks!

The mimrgns command is unfortunately not working. STATA is giving the following error:
command mimrgns is unrecognized
r(199);

I do not want that the logistic regression is using one of the countries as reference, but that the weighted mean incidence is used as the reference.

Best,
Anna

[daniel klein]

As indicated, you have to install mimrgns from the SSC. That is, you have to type

ssc install mimrgns

once, before you can use it.

I do not want that the logistic regression is using one of the countries as reference, but that the weighted mean incidence is used as the reference.

In that case, I guess you need to look up the respective coding and generate the indicators (in each of the imputed datasets) yourself.

Best
Daniel

[Anna Seijmonsbergen]

Okay, thank you!