Hello everyone,
My name is Adnan Fakir and I am working on a binary response model with a discrete endogenous regressor. I am estimating the model using Lewbel's special regressor procedure, using Baum's written sspecialreg command in Stata 12. I am however having difficulties in conducting the postestimation tests for instrument validity, such as as Anderson's under-identification test and Sargan-Hansen's over-identification test. Does performing the usual estat overid after ssimplereg give reliable test results or is there another recommended way to do this?
Bontemps and Nauges (2014) do this in their working paper (Link: https://www.tse-fr.eu/publications/i...drinking-water) but they mention in a footnote that: "These tests for instrument validity were incorporated by the authors into the sspecialreg procedure (Baum, 2012). The updated procedure is available upon request." This makes me think that the usual way of performing these tests is not valid.
My command in Stata is as follows:
sspecialreg PIPVBin mheight, trim(5) kdens exog(wpublic mcontrol knowind usedmr interIPV wage agesq agediff wmarrage wompri womsec womhigh edugap wislam wurban wpoorer wmiddle wricher wrichest HHmemb wserv wagri wentre semiskill wprof barisal ctg khulna rajshahi sylhet rangpur) endog (auto) iv(exfam wifesig) hetero
I will be more than happy to post my results and details of the dataset I am using if that is needed.
Looking forward to your replies.
Best,
Adnan
My name is Adnan Fakir and I am working on a binary response model with a discrete endogenous regressor. I am estimating the model using Lewbel's special regressor procedure, using Baum's written sspecialreg command in Stata 12. I am however having difficulties in conducting the postestimation tests for instrument validity, such as as Anderson's under-identification test and Sargan-Hansen's over-identification test. Does performing the usual estat overid after ssimplereg give reliable test results or is there another recommended way to do this?
Bontemps and Nauges (2014) do this in their working paper (Link: https://www.tse-fr.eu/publications/i...drinking-water) but they mention in a footnote that: "These tests for instrument validity were incorporated by the authors into the sspecialreg procedure (Baum, 2012). The updated procedure is available upon request." This makes me think that the usual way of performing these tests is not valid.
My command in Stata is as follows:
sspecialreg PIPVBin mheight, trim(5) kdens exog(wpublic mcontrol knowind usedmr interIPV wage agesq agediff wmarrage wompri womsec womhigh edugap wislam wurban wpoorer wmiddle wricher wrichest HHmemb wserv wagri wentre semiskill wprof barisal ctg khulna rajshahi sylhet rangpur) endog (auto) iv(exfam wifesig) hetero
I will be more than happy to post my results and details of the dataset I am using if that is needed.
Looking forward to your replies.
Best,
Adnan