Jonah B. Gelbach's Stata ado-Files Page

Please note: all code is provided for convenience to other user, but NO PROMISE OF CORRECTNESS is made! You may want to check back for updates/bug fixes/etc.

• File to cut down on code-writing for merges

  • jmerge.ado (Executes the program; syntax at top of file)
  
  
  

• File to run difference-in-differences

  • did2.ado (Executes the program; syntax at top of file)
  
  
  

• Files to run fixed effects 2SLS (based on areg)

  Changelog:
  1. Note: problem with expanding lists of variables has been fixed
  2. On this web page and in an Oct 17, 2000 posting to the statalist, I wrongly said the following:

     Standard errors were previously incorrect, as the standard error of the regression was being computed using fitted rather than actual data for endogenous variables. This problem was fixed on October 17th, 2000.

     This was wrong because the original code was actually doing it right. I've restored the code to its original form, aside from a comment explaining why changing the names of the temp matrices `b' and `V' ensures that the correct standard errors are generated.

     • areg2.ado (Executes the program)
     • areg2.hlp (Help file)
     
     
     

  3. Files to calculate mean marginal effects for probit and logit models
     The current version is 4.3 and was posted on August 27, 2004. This version fixed a couple bugs in v4.1/4.2. It should work fine on all fronts.)

     • margfx.ado (Executes the program)
     • margfx.hlp (Help file)
     • Documentation (please read this rather than emailing me):
       • PDF Format
     
     
     

  4. Files to calculate z-statistics on the difference of two independent statistics given their standard errors

     • sigdif.ado (Executes the program)
     • sigdif.hlp (Help file)
     
     
     

  5. File to calculate covariance matrix for a set of variables

     • covar.ado (Executes the program)
     
     
     

  6. Files to run 2-step tobit and probit with endogenous rhs variables.

     • Please note: the code for these programs will execute for any rhs endogenous variable. However, the estimates are not consistent when the rhs endogenous variables are not continuous. The distribution theory is developed in
       
       
       • Newey, Whitney K., "Efficient Estimation of Limited Dependent Variable Models with Endogenous Explanatory Variables" Journal of Econometrics, v36 n3 November 1987, pp. 231-50.
         
         
       To see more on this issue, please click here.
       
       
     • A shortcoming of the code in these files is that they do not compute consistent standard errors. The files estimate "second-step" tobit/probit models with predicted values of the endogenous RHS variable (and the first-stage prediction error -- see Newey's article on this). For the usual reasons, the standard errors that Stata reports for these second-step estimates are incorrect, because they do not correct for the first-stage prediction. If you are worried about this issue, you will have to correct the standard errors yourself -- Maddala's textbook is a fine reference on how to do so. Newey and McFadden also have a paper in the Handbook of Econometrics series describing the issues and the correction.
       
       
     • Also, please understand that I am providing access to these files to be helpful, and I get many inquiries about these files. I do not have enough time to answer them promptly or in any detail. Please do not ask me to interpret the output or elaborate in great detail. The code is fairly well documented, so you should look at it to figure out what's going on.
       
       
     • Lastly, please note that the theory developed in Newey's article requires joint normality of all unobserved terms in the full model. I do not know anything about "logit IV", or other such plausible extensions. I also do not know whether ordered probit can be done this way. Because I virtually never use these estimators, I have no impending plans to find out. These files are what I have to offer, so if you want something else you will have to figure it out on your own. However, one last option is to check out the -probiv- files written by Deon Filmer of the World Bank. You can get these files here, though neither Deon nor I can support them.
       
       
     • tobitiv.ado (Executes the program)
     • probitiv.ado (Executes the program)
     • Example of output and discussion of interpretation (applies to both -probitiv- and -tobitiv-).
     
     
     
  7. State Fips Codes (brought to you by Diane Whitmore): statename.do