    # st: Cumulative response functions

 From Felipe Balmaceda To statalist@hsphsun2.harvard.edu Subject st: Cumulative response functions Date Wed, 15 Mar 2006 13:36:02 -0400

Dear all,

I am trying to obtain cumulative response functions (CRFs). In order to do so I wrote the following program

quietly xtreg dgas enap1 gas1 tiempo inc des inc1-inc6 des1-des6 gasinc1 gasdes1,fe robust cluster(est)

predictnl LR = (_b[enap1]/_b[gas1])*(-1), se(seLR) ci(LLR ULR)
predictnl B0 = (_b[inc]), se(seB0) ci(LB0 UB0)
list B0 seB0 LB0 UB0 in 1
predictnl B1 = B0+_b[inc1]+_b[gasinc1]*B0+_b[gas1]*(B0-LR), se(seB1) ci(LB1 UB1)
list B1 seB1 LB1 UB1 in 1
predictnl B2 = B1+_b[inc2]+_b[gasinc1]*(B1-B0)+_b[gas1]*(B1-LR), se(seB2) ci(LB2 UB2)
list B2 seB2 LB2 UB2 in 1
predictnl B3 = B2+_b[inc3]+_b[gasinc1]*(B2-B1)+_b[gas1]*(B2-LR), se(seB3) ci(LB3 UB3)
>
>
>
predictnl B25 = B24+_b[inc3]+_b[gasinc1]*(B24-B23)+_b[gas1]*(B24-LR), se(seB25) ci(LB25 UB25)

The problem is that predictnl (nlcom) takes B0, B1 and so and so forth as fixed number and not as combination of parameters with a given distribution and thus the standard errors obtained by the delta method are not the correct ones. If If I enter the whole expressions, stata 9 complains that the expressions are too long and provides no results.

Does someone can suggest me a way of obtaining the CRFs (being new in stata, I run out of ideas)

thanks to all

felipe

*
* 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/