Hello,
I have been following some of the discussions in other posts, but have had a hard time finding the answer to my issue.
I see from other users that xtivreg2 provides post estimation tests for fe and fd models, but not re models. I cannot use fe or fd because some of my variables are time invariant.
I use the following code to check for over identification as per a previous post:
- reg l_impvalUSD conn_after ref_2008 $controls
- predict v2hat, resid
- xtreg l_impvalUSD connected v2hat conn_after ref_2008 $controls B1_elect_gift, re vce(cluster id)
This gives me a highly significant P score for the z statistic, which from my understanding, would indicate that my y2 (connected) is endogenous.
The next step for me is to find out :
1. how to manually check for weak/strong instruments for the case where I use one instrument (as in here I use only B1_elect_gift). Would you have advice on how to do this when in an re setting? with xtivreg2 , fe or fd , we have the test already done for us, but not with xtivreg , re
2. Would you similarly have advice on how to test for overidentification when I have more than one instrumental variable when in an xtivreg ,re model?
Here is an except from my datafile.
I have been following some of the discussions in other posts, but have had a hard time finding the answer to my issue.
I see from other users that xtivreg2 provides post estimation tests for fe and fd models, but not re models. I cannot use fe or fd because some of my variables are time invariant.
I use the following code to check for over identification as per a previous post:
- reg l_impvalUSD conn_after ref_2008 $controls
- predict v2hat, resid
- xtreg l_impvalUSD connected v2hat conn_after ref_2008 $controls B1_elect_gift, re vce(cluster id)
This gives me a highly significant P score for the z statistic, which from my understanding, would indicate that my y2 (connected) is endogenous.
The next step for me is to find out :
1. how to manually check for weak/strong instruments for the case where I use one instrument (as in here I use only B1_elect_gift). Would you have advice on how to do this when in an re setting? with xtivreg2 , fe or fd , we have the test already done for us, but not with xtivreg , re
2. Would you similarly have advice on how to test for overidentification when I have more than one instrumental variable when in an xtivreg ,re model?
Here is an except from my datafile.
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input float(l_impvalUSD connected) byte conn_after float(ref_2008 B1_elect_gift) 11.809468 . . 1 .75 7.6316 . . 1 .14285715 9.741706 . . 1 .3333333 10.00367 . . 1 . 5.265005 . . 0 . 9.185537 . . 1 . 12.123355 . . 1 . 5.430605 1 1 1 .4545455 6.105125 1 1 1 .4545455 9.962399 . . 1 . 6.369867 0 0 1 . 9.931567 . . 1 . 10.376081 . . 1 . 6.545316 . . 1 .14285715 5.254809 . . 1 . 15.86338 1 1 1 . 13.419892 1 1 1 . 7.630236 . . 1 .5 8.558739 . . 1 0 8.743931 . . 1 .5 8.670208 . . 0 0 8.994358 . . 1 .5 13.358562 1 1 1 . 9.680139 . . 1 .14285715 10.097457 . . 0 0 4.3324513 . . 1 .5 10.08326 . . 1 . 8.90372 . . 1 . 9.158985 . . 1 0 7.087101 . . 1 . 7.519593 1 1 1 .4545455 10.39248 . . 1 0 9.862251 . . 1 0 7.956739 . . 1 0 4.4641743 . . 0 0 9.229189 . . 1 .14285715 13.067918 . . 1 .5 12.750822 . . 1 0 7.678548 1 0 0 .4545455 4.723966 . . 1 . 3.975444 . . 1 . end label values connected CON label def CON 0 "not connected", modify label def CON 1 "connected", modify
Comment