Thanks Nichols, Nick, and Mark. Nichols got my question (E[Y|x1,x2,con=1] =
E[Y|x1,x2,con=0]). Y1 and Y2 are different. Let me tell you the whole story
in short as follow.
I am assessing the effect of soil conservation technology on crop yields
(productivity) using plot level data. To do this I am following Fuglie and
Bosch�s (1995) and Khanna�s (2001) work. Their work is similar to Lee�s
(1978). They were estimated separate regression for technology adopters and
non-adopters. From the regressions they estimated fitted values at
representative values (mean) of some of the regressors and then they
compared the mean predicted values. They used endogenous switching
regression analysis.
What I posted is exogenous switching regression model. In my case as in the
above authors, I divided the sample into those plots that adopt soil
conservation technlogy and those that do not. In the regressions below the
variable con== 1 indicates those who adopted the technology and con == 0
those that do not. The variable y1 and y2 also indicate yields obtained from
adopters and non-adopter of soil conservation technology, respectively.
areg y1 x1 x2 if con ==1, absorb(hhno) cluster(hhno)
predict conhat if con ==1, xbd
areg y2 x1 x2 if con ==0, absorb(hhno) cluster(hhno)
predict wconhat if con ==0, xbd
Hope this will help. Thanks again for your help.
Dawit
Fuglie, K. O., and Bosch, D. J. 1995. Economic and Environmental
Implications of Soil Nitrogen Testing: A Switching�Regression Analysis.
American Journal of Agricultural Economics 77: 891- 900.
Khanna, M. 2001. Sequential adoption of site-specific technologies and its
implication for nitrogen productivity: A double selectivity model. American
Journal of Agricultural Economics 83(1): 35-51.
Lee, L.F. 1978. Unionism and wage rates: A simultaneous equations model with
qualitative and limited dependent variables. International Economic Review
19(2):415-433.
From: "Nichols, Austin" <[email protected]>
To: 'Mark Schaffer' <[email protected]>, "'[email protected]'"
<[email protected]>, "'[email protected]'" <[email protected]>
Subject: RE: st: RE: test predicted values
Date: Sun, 16 Jan 2005 15:44:41 -0500
I think that Dawit Lidia wants to test that
E[Y|x1,x2,con=1] = E[Y|x1,x2,con=0]
where Y=(y1, y2) and it's unclear whether y1=y2 from the post.
-----Original Message-----
From: Nichols, Austin
Sent: Sunday, January 16, 2005 3:37 PM
To: '[email protected]'
Cc: '[email protected]'
Subject: RE: st: RE: test predicted values
Are the dependent variables in the 2 models really different vars?
If not, Nick has indicated the solution:
. g double x1c1=x1*con
. g double x2c1=x2*con
. areg y x1 x2 con x1c1 x2c2, absorb(hhno) cluster(hhno)
. test con x1c1 x2c2
will test for difference in predicted vals across the 2 models.
-----Original Message-----
From: Dawit Lidia [mailto:[email protected]]
Sent: Sunday, January 16, 2005 3:09 PM
To: [email protected]
Subject: RE: st: RE: test predicted values
Dear Nick Cox,
Thanks you for your response. I think I have to refine my question as
follow. Can I do the following test on predicted values without adjusting
the standard errors since observations are not independent within group
(hhno)?
ttest conhat = wconhat unequal
where conhat and wconhat are predicted below.
areg y1 x1 x2 if con ==1, absorb(hhno) cluster(hhno)
predict conhat if con ==1, xbd
areg y2 x1 x2 if con ==0, absorb(hhno) cluster(hhno)
predict wconhat if con ==0, xbd
Hope this will be clear. Many thanks for your help.
Dawit
>From: "Nick Cox" <[email protected]>
>Reply-To: [email protected]
>To: <[email protected]>
>Subject: st: RE: test predicted values
>Date: Sun, 16 Jan 2005 16:35:28 -0000
>
>It would seem more straightforward,
>and more appropriate, to fit
>
>areg y1 x1 x2 con, absorb(hhno) cluster(hhno)
>
>which provides your test directly -- unless
>I am missing some subtlety here.
>
>Nick
>[email protected]
>
>Dawit Lidia
>
> > I am running the following types of regressions
> >
> > areg y1 x1 x2 if con ==1, absorb(hhno) cluster(hhno)
> > predict conhat if con ==1, xbd
> >
> > areg y2 x1 x2 if con ==0, absorb(hhno) cluster(hhno)
> > predict wconhat if con ==0, xbd
> >
> > y1 and y2 are dependent variables and xs are independent variables.
> >
> >
> > I want to test if there is mean difference between the two
> > predicted values
> > uisng simple 'ttest'.
> >
> > My question is, do I need to correct standard errors(se) of
> > the 'ttest'
> > since the observed ys are not independent. If the answer is
> > yes, would you
> > please suggest me how i can correct it or any other
> > altenative test that can
> > adjust se.
>
>*
>* 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/
_________________________________________________________________
MSN Hotmail http://www.hotmail.com Med markedets beste SPAM-filter. Gratis!
*
* 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/
_________________________________________________________________
MSN Hotmail http://www.hotmail.com Med markedets beste SPAM-filter. Gratis!
*
* 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/
