Margins after xtgee

[Chiara Piazzo]

Dear all,

I am running a GEE regression:

xi:qic access var1 var2 var3, i(city) family(gauss) link(identity) corr(unstructured) robust

All the explanatory variables are continuous and in logs while the dependent variable. I was wondering if wanted elasticities can I do the following:

margins, eyex(var1 var2 var3) atmeans

Thank you!

[Tom Weichle]

Hi Chiara,

Your title of the post and the description of what you plan to do are a little confusing.

From my understanding, -qic- is a user-written command which can perform GEE analyses and also produces the QIC criterion to select the best-working correlation structure. However, -qic- is an older command which does not support Stata's -margins- command. So, after comparing the criterion measures of different correlation structures produced from -qic-, you can select the model and correlation structure to run in -xtgee-. Then, you can use the -margins- command to calculate the elasticities.

Hope this helps!

Tom

[Chiara Piazzo]

Dear Tom,

Thank you for your reply. Apologies for he mistake, I forgot to paste the -xtgee- command after using -qic- to select the appropriate correlation structure.

xtgee access var1 var2 var3, i(city) family(gauss) link(identity) corr(unstructured) robust

Then the following should give me the average marginal effects of all covariates

margins, dydx(*)

But I am interested in elasticity and my specification is linear-log, so should I be using -dyex- to compute the elasticities? Or should it be

margins, dydx(var1 var2 var3)

Apologies for not providing all the information in the original post? Thanks again!

[Tom Weichle]

Hi Chiara,

From the -margins- help file, you can produce the following:

dydx(varlist) estimate marginal effect of variables in varlist
eyex(varlist) estimate elasticities of variables in varlist
dyex(varlist) estimate semielasticity -- d(y)/d(lnx)
eydx(varlist) estimate semielasticity -- d(lny)/d(x)

dydx(varlist), eyex(varlist), dyex(varlist), and eydx(varlist) request that margins report derivatives of the response with respect to varlist rather than on the response itself. eyex(), dyex(), and eydx() report derivatives as elasticities; see Expressing derivatives as elasticities in [R] margins.

Tom

[Chiara Piazzo]

Tom Weichle - thank you very much!