Dear Statalisters,
I am trying to implement a procedure proposed by Hausman, Abrevaya and
Scott-Morton to control for misclassification in the dependent variable
in a discrete-response setting. Below is the code I have been using but
I am unable to program up a condition needed for identification, namely
that the sum of the two misclassification probabilities, a0+a1, is less
then one. I think this condition guarantees that the signs of the
parameters are identified as well as their magnitudes. When I run the
program every estimate comes out with the opposite sign to the estimates
from a standard probit model. I think I need to reparameterise the
misclassification probabilities, a0 and a1, in some way but so far I
have been unsuccessful. Any suggestions?
capture program drop hausprob
program define hausprob
args lnf b a0 a1
tempvar p
quietly gen double `p'=normal(`b')
quietly replace `lnf'=ln(`a0'+(1-`a0'-`a1')*`p') if $ML_y1==1
quietly replace `lnf'=ln(1-`a0'-(1-`a0'-`a1')*`p') if $ML_y1==0
End
Many Thanks,
Adele
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
