Postestimation xtabond "estat abond"
Hello,
I am currently working on my thesis I am using the xtabond command. I have to use 12 regressions ir order to test my hypothesis: the impact of information and communication technologies is positive and different in states with different income levels. Using postestimation method "estat abond" the results of 3 of the 12 are between .5 and .10. I do not know if I should stay with that result and say that the hypothesis of autocorrelation at 5 percent is not rejected or I have to change something in my specification of the model.
My principal regression is xtabond LOGPIBPC INT MOV FIJ CREC TRABCONT PREP SALUD IED FIN, noconstant vce(robust). INT MOV and FIJ are my main variables and I alternate them to see the effect of fixed, mobile and internet technology separately. I repeat the same process for two subsamples per income "high" and "low". The problems arise in the whole sample when only the variable MOV is present and for the low income subsample when only INT is present and the same for MOV. I do not know whether to change the specification or if it is acceptable to say that I do not reject the aforementioned hypothesis at 5 percent.
Thank you
I am currently working on my thesis I am using the xtabond command. I have to use 12 regressions ir order to test my hypothesis: the impact of information and communication technologies is positive and different in states with different income levels. Using postestimation method "estat abond" the results of 3 of the 12 are between .5 and .10. I do not know if I should stay with that result and say that the hypothesis of autocorrelation at 5 percent is not rejected or I have to change something in my specification of the model.
My principal regression is xtabond LOGPIBPC INT MOV FIJ CREC TRABCONT PREP SALUD IED FIN, noconstant vce(robust). INT MOV and FIJ are my main variables and I alternate them to see the effect of fixed, mobile and internet technology separately. I repeat the same process for two subsamples per income "high" and "low". The problems arise in the whole sample when only the variable MOV is present and for the low income subsample when only INT is present and the same for MOV. I do not know whether to change the specification or if it is acceptable to say that I do not reject the aforementioned hypothesis at 5 percent.
Thank you
