Yes! After estimating the model, use predict X, reffects (where X is any name you want to give it). 

I hope this helps

Ariel

I forgot to mention that the ,reffects option will save all the estimates for all coefficients and levels in your model. So I'd use something like predict re*, reffects if there is more than one RE estimate in your model.

Thanks Ariel for your quick response!

This way I get individual predictions for the random effects. What I would like to see are regression coefficients for each stage of my random effects variable, just as I get them for each categorical fixed effects variable.

Let me give you some more details: I would like to estimate age, period, and cohort (APC) effects in a sample of 7 repeated cross-sectional surveys by means of cross-classified random effects (CCREM) models, as proposed by Yang & Land (Yang, Y., Land, K.C., 2008, see attached paper). In line with this, I model age and other (categorical) covariates as fixed effects on level 1, while I include period (7 surveys) and cohort (11 5-year groups) as random effects on level  2.

As a result of e.g. mixed or meqrlogit in Stata I get one estimator for each of the level 2 variables, which indicate the variation in the dependent variable by period and cohort (= variance components). However, I want to have a random effects coefficient for each period k (and for each cohort), that is, the contribution of period k averaged over all cohorts. So in my example I should get a total of 7 coefficients for period and 11 coefficients for cohort, in addition to the variance components. Please also see the attached paper (table on page 315).

Thanks again for the help!

Best,

Alex

You may be interested in these articles  on random effects APC model - you will see that we have concerns

we used RunMlwin from within Stata because of the cross classified nature of the structure of the APC model

http://www.bristol.ac.uk/cmm/software/runmlwin/

But I do not know enough about 'standard' Stata's capabilities  in handling non-hierarchical structures

https://www.researchgate.net/publication/236217148_Another_%27futile_quest%27_A_simulation_study_of_Yang_and_Land%27s_Hierarchical_Age-Period-Cohort_model

https://www.researchgate.net/publication/256326448_Don%27t_birth_cohorts_matter_A_commentary_and_simulation_exercise_on_Reither_Hauser_and_Yang%27s_%282009%29_age-period-cohort_study_of_obesity

https://www.researchgate.net/publication/236217153_Current_practice_in_the_modelling_of_Age_Period_and_Cohort_effects_with_panel_data_a_commentary_on_Tawfik_et_al._%282012%29_Clarke_et_al._%282009%29_and_McCulloch_%282012%29

https://www.researchgate.net/publication/271134696_Should_age-period-cohort_analysts_accept_innovation_without_scrutiny_A_response_to_Reither_Masters_Yang_Powers_Zheng_and_Land

Article Another 'futile quest'? A simulation study of Yang and Land'...

Article Current practice in the modelling of Age, Period and Cohort ...

Article Don't birth cohorts matter? A commentary and simulation exer...

Article Should age-period-cohort analysts accept innovation without ...

It is possible to run cross level models in Stata, but I am not sure if it is possible to retrieve those specific coefficients. I'll play around with it and see if I can figure it out.

Thanks Kelvyn for pointing out the limits of this approach - this is very interesting and I will have a look on the attached papers.

Ariel, thank you for your help. As I understand it, the R.varname option for specifying random effects does exactly what I want: It handles the levels of varname as a series of indicator variables. Unfortunately it seems that Stata does this only internally without saving those estimates in the e() container.

Dear Alexander Pabst

May be after running your program, mat list r(table) works. If so just specify which element of table do you like to save, for example by comment

 scalar B1age=el(r(table),1,2)

element of first row and second column would be save in B1age

 I hope this helps

Amir Hamta