[Niels Henrik Bruun]

Hi
Been on weekend. Since saturday there has been a new update at ssc such that -ic- works with -stcox-
Tiago Pereira: For now you have to use -ici-. In the -ic- help there is an example describing how -icp- works: Prefix icp and how it works.
The steps there are the ones to do when using -glm-

[Nicholas Graham]

Once again: many thanks!

[Marianne Holm]

Dear Niels Henrik,

Sorry for the late reply. I have now re-installed the updated package and it runs the analyses in the Cox regressions perfectly.
Thanks again very much for developing this package.

Can I ask how I best reference/acknowledge you in my publications?

Kind regards,
Marianne Holm

[Niels Henrik Bruun]

Hi Marianne
Look at Export references at https://econpapers.repec.org/softwar...de/s457975.htm
There are references in Bibtex, Ris and different versions of html.

Thank you for acknowledging our work.

[Lisa van Zutphen]

Hello!

I was wondering if the package runs on mi data? And if it should run on mi data; what code or commands should I use?

Thank you!

[Niels Henrik Bruun]

Hi Lisa
I think you can use -ici- with the returned vector e(b_mi) and matrix e(V_mi).
You have to generate your dummy-variables yourself as described in the examples in the help.
I would generate the dummies before imputation and use them in the imputation and the models afterwards.

Kind regards /nhb

[James Oguttu]

Hi Niels,
I am able to test for interaction using the ic command in stata using the unadjusted model. My question however is how could I adopt the ic command on a full model with 6 explanatory variables when i interested in testing for interaction between only two of the explanatory variables. Secondly, is it possible to test for interaction when fitting multilevel models?

[Niels Henrik Bruun]

Hi James
I'm not shure what you mean here:

I am able to test for interaction using the ic command in stata using the unadjusted model.

.

Regarding

My question however is how could I adopt the ic command on a full model with 6 explanatory variables when i interested in testing for interaction between only two of the explanatory variables.

the -IC- command and setup works only for 3 dichotomous variables of which one is the outcome.
My immediate suggestion is to analyse pairs of dichotomous variables explaining a dichothomous outcome.

Finally, the commands -ic- and -icp- has not been set up for multi level models.
It should be possible to use the -ici- command for a multi level regression, but you should be very careful about setting the regression used to deliver the estimates and their variance matrix.
Look in the help for guidance.
The rule "garbage in garbage out" must certainly apply here.

[Halvor Sommerfelt]

Dear Niels Henrik:

I am doing a logistic regression on a case-control dataset (so estimating odds ratios), i.e. with a dichotomous outcome (Y) and a dichotomous factor (my exposure, A) but with the second factor (the stratifying variable, B) having three categories. Does the ic(p) command handle such a situation? If so, how do I interpret the two RERIs I expect to get?

With best regards from

Halvor Sommerfelt

[Niels Henrik Bruun]

Dear Halvor
The articles behind -ic- command only handle 2 by 2 by 2 tables.
By construction the theory is not extendable.
Kind regards
nhb

[Hillel Alpert]

Does icp work with mixed effects models? The command icp, show: mepoisson, etc . . .runs without random effects terms (e.g. || level1: ||level2) but not with these terms.

Does it work with svy: commands?

[Niels Henrik Bruun]

Dear Hillel
The ic should be able to run with mixed effect models.
What kind of error do you get?

[Nerea Becerra]

Hi,
could some someone tell me why this command cannot be used after clogit?

Thank you

[Niels Henrik Bruun]

Dear Nerea
Probably because the theory in the reference articles did not consider it as I recall.
I was afraid of promising to much.
However, -icp- works with mixed models.
So until an update appears, consider refrasing your model in melogit instead of clogit.
Kind regards
nhb

[Nerea Becerra]

Thank you for your prompt reply, Niels.
I think I will wait for an update since I am conducting a case-control study.

Kind regards,

Nerea