I'm applying the control function (CF) approach in a nonlinear model using a panel dataset of five waves (wave 1 - 5). I have run a fixed effect and random effect, the Hausman test revealed that the fixed effect is accepted.
Then, I want to do a robust check of my result - I decided to run 2SLS, but my regressor is not statistically significant in the second stage. I attempted the control function approach, which is the alternative to 2SLS. First, it was statistically significant, but when I run it again it is not statistically significant.
Can someone help check if I have missed any step, and what method can I use to do a robustness check of fixed effect? I have tried GMM, but it was not making sense.
Kindly note: both the dependent variable and regressor are continuous, even IV (diffage - the difference between the age of male and female in the household).
2SLS
First stage
reg Autonomy4 diffage age sex_c cageN hhsize i.femaleedu hhincome healthdum Qlifedum emplstatusdum i.religion dwel african indian white parent grandparent uncle_ant
estat endog
estat firststage
Second stage
**** 2SLS
ivreg2 stunting age sex_c cageN hhsize i.femaleedu hhincome healthdum Qlifedum emplstatusdum religion dwel african indian white parent grandparent uncle_ant (Autonomy4 = diffage)
or
ivregress 2SLS stunting age sex_c cageN hhsize i.femaleedu hhincome healthdum Qlifedum emplstatusdum religion dwel african indian white parent grandparent uncle_ant (Autonomy4 = diffage), first
Control Function Approach
First stage
reg Autonomy4 diffage age sex_c cageN hhsize i.femaleedu hhincome healthdum Qlifedum emplstatusdum i.religion dwel african indian white parent grandparent uncle_ant
test diffage
predict error1hat, residual
Second stage
ivreg2 stunting Autonomy4 age sex_c cageN hhsize i.femaleedu hhincome healthdum Qlifedum emplstatusdum i.religion dwel african indian white parent grandparent uncle_ant error1hat
To show the effect: I specified below graph or plot
margins, at(Autonomy4=(-1.02(0.01)1)) atmeans
marginsplot
Thanks for your help.
0 Response to Control Function Approach and looking for a method to do robustness check of fixed effect
Post a Comment