Dear STATALIST,

I am running a Generalized Estimating Equation in STATA with link(log) corr(independent) family(gamma) and vce(robust). The results of the command...
Code:
 
 xtgee dependentvariable independentvariable controlvariables, link(log) corr(independent) iterate(100) family(gamma) vce(robust)
...provides me with a negative value for the beta coefficient

According to Ballinger (2004) "The regression coefficients that result from GEE models for logit, probit, and log links need to be exponentiated before they are meaningful." I am using a log link. If I get this the right way then the -eform- (according to the STATA help displays the exponentiated coefficients and corresponding standard errors and confidence intervals) option should be used for this.

So I am running:
Code:
 
 xtgee dependentvariable independentvariable controls, link(log) corr(independent) iterate(100) family(gamma) vce(robust) eform
But now the value for the beta coefficient turns out to be positive. Is it correct to only interpret the positive beta coefficient after adding -eform-?