Hi Sachin,
Unfortunately my data has an extremely high level of censorship, often
around 70 in some subsamples. As a result, it appears that the literature
indicates that a pure CLAD (i.e., one focused solely on the 0.50 quantile) would
not be appropriate for my data. That is why I had not yet contacted
Dr. M. O. Wilhelm to see if he would be willing to share his -CLADTEMP- program.
Rather, I concentrated my efforts on the CQR methodology. I did
attempt to use the Ken Chay ado file, however, it failed to converge. Then
I turned my attention to the ado file written by Robert Vigfusson incorporating
the Buchinsky (1991, 1994) algorithm which is available from Timothy G.
Conley at: http://faculty.chicagobooth.edu/timothy.conley/research/qrcode/quantile.html.
Once again, there were several of the quantiles that did failed to converge.
I attempted to overcome this by modifying the code per Fitzenberger (1994)
suggestion of keeping observations that were either equal to or greater than
zero rather than solely those that were greater than zero. Once again,
convergence was still a problem.
Fitzenberger (1994). A note on estimating censored quantile regressions,
Discussion Paper 14, University of Konstanz.
Buchinsky (1994). Changes in the U.S. wage structure 1963-1987: Application
of quantile regression, Econometrica 62 (2), 405-458.
Cameron McIntosh had kindly responded earlier to check out
a recent work by Chernozhukov and Hansen:
Chernozhukov, V., & Hansen, C. (2008). Instrumental variable quantile regresssion: A robust inference approach. Journal of Econometrics, 142(1), 379-398.
http://faculty.chicagobooth.edu/christian.hansen/research/CH_IQR_Aug2_06.pdf
I have considered attempting to contact Hansen. My main concern is that
I do not have any variables that appear meet the exclusion restriction (i.e.,
strongly correlated with the selection equation but not with the final model
equation) that is required for instrumental variables.
I am currently working attempting to write up a code for Chernozhukov
and Hong (2002). As Cameron had thought, the macro for the selection
has been a major point of difficulty.
Chernozhukov, V., & Hong, H. (2002). Three-step censored quantile regression
and extramarital affairs. Journal of The American Statistical Association, 97(459), 872-882.
Once again, thank you to those who have responded and I of course would be thankful
for any further suggestions.
Also, Sachin, I would certainly be willing to discuss this further if you are interested.
Steven
>>> Sachin Chintawar <[email protected]> 08/11/09 11:45 AM >>>
Hi Mr. Steven
I have been working on the same problem since six months. While I did
not find any ado file I am trying to write one of my own. In the
meantime here is some of the things that I have been following that
could help you in your research (All the three are based on a 2 step
estimation method and try to answer the convergence problem)
A few things that you may want to know. Dr. Dean Jolliffe has
contributed to devloping an official CLAD procedure. Please check for
that using
-help clad-
Dr. Ken Chay's ado file. is posted o his webite. Type -help cqr- to
find his work as well has his paper where he uses it.
Another procedure defined by Dr. M. O. Wilhelm who has written a user
defined procedure -CLADTEMP- You can refer to his paper
Wilhelm, Mark Ottoni (2008). Practical Considerations for Choosing
Between Tobit and SCLS or CLAD Estimators for Censored Regression
Models with an Application to Charitable Giving. Oxford Bulletin of
Economics and Statistics 70(4): 559-582.
I hope this helps. I would like to share notes with you f you do find
an available ado file or if mine works perfectly
.
Best of luck with your program
Sachin
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
</[email protected]>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
