I'm having trouble bootstrapping standard errors of a coefficient vector from a 2SLS regression, where the first stage is mlogit with 5 instrumental variables, put into a second stage linear regression equation. The code goes like this:
capture program drop proca_reg
program define proca_reg, eclass
tempvar p1 p2 p3 p4 p5 *the instrumental variables
tempname coefs
mlogit treatment age age65plus northeast midwest west year hhinc99 married urban preyear_exp_def ncicombined_index urologypop3d radoncpop3d imrtpop brachypop *1st stage
predict `p1', outcome(1) *predicted values for instrumental variables
predict `p2', outcome(2)
predict `p3', outcome(3)
predict `p4', outcome(4)
predict `p5', outcome(5)
regress totyear1_2_exp_def `p1' `p2' `p3' `p4' `p5' age age65plus northeast midwest west year hhinc99 married urban preyear_exp_def ncicombined_index *2nd stage
mat def `coefs' = e(b)
ereturn post `coefs'
ereturn local cmd = "bootstrap"
end
bootstrap "proca_reg" _b *bootstrap coefficients coming from the above program
When I used the above code for the first run, I got coefficients for ALL variables but I only got the bootstrapped standard errors for the regressors that are NOT the instrumental variables. It wouldn't bootstrap standard errors for coefficients for p1, p2, p3, p4, p5. The STATA error code says something about missing observations or values.
**Does anyone know what is going on here? Or better yet, how to fix this?
Also, when I repeated the code, it won't generate anything, it stops staying "_000006 already define". I looked around the dataset and there are now several dummy variables that I think correspond to the predicted outcomes for the instrumental variables, which the program or code isn't clearing each time the bootstrap is run without saving. Also, any ideas how to fix this?
Thanks for your thoughts
Daniella
*
* 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/
