Thank you everyone for all the suggestions. I was able to tweak Thomas's
code and get the result I pretended (see code below):
As for the use of the jacknife estimator, I'm not sure if it is doing the
correct estimation (or at least what I need). For rep78=3 and just for the
first observation I get:
sysuse auto, clear
keep if rep78==3
gen i=_n
jacknife _b, keep: reg mpg weight displacement turn
reg mpg weight displacement turn if i!=1
Regress reports the following coefficients:
----------------------------------------------------------------------------
--
mpg | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
weight | -.0020761 .0016688 -1.24 0.225 -.005513
.0013608
displacement | -.0062576 .0119518 -0.52 0.605 -.0308729
.0183577
turn | -.3671771 .1733158 -2.12 0.044 -.7241277
-.0102266
_cons | 42.77554 5.130264 8.34 0.000 32.20957
53.34152
----------------------------------------------------------------------------
--
While the results for using jacknife for the first observation are:
_b_weight _b_displacement _b_turn _b_cons
.0009151 -.038999 -.3450164 40.28076
After taking a look to the manual, it seems that jacknife calculates
pseudovalues, which isn't what I need. I just need to use the information of
the rest of the firms on the same SIC code (excluding the company being
analyzed) to calculate the expected value of the company being analyzed.
************* Begin Code *************
sysuse auto, clear
gen mpg_excl=.
gen cweight=.
gen cdisplacement=.
gen cturn=.
gen ccons=.
foreach n of numlist 2/5 {
preserve
drop if rep78!=`n'
qui gen i=_n
qui sum mpg
local nobs=r(N)
foreach j of numlist 1/`nobs' {
tempvar a b c d
qui regress mpg weight displacement turn if i!=`j'
qui predict hat
qui replace mpg_excl = hat if i == `j'
qui drop hat
qui gen `a'=_b[weight]
qui gen `b'=_b[displacement]
qui gen `c'=_b[turn]
qui gen `d'=_b[_cons]
qui replace cweight=`a' if i==`j'
qui replace cdisplacement=`b' if i==`j'
qui replace cturn=`c' if i==`j'
qui replace ccons=`d' if i==`j'
drop `a' `b' `c' `d'
}
qui gen
est=ccons+cweight*weight+cdisplacement*displacement+cturn*turn
save c:\file`n'.dta, replace
restore
}
************* End Code *************
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Steichen, Thomas
J.
Sent: sexta-feira, 6 de Julho de 2007 21:34
To: [email protected]
Subject: st: RE: Running a regression for a subset of observations issue
Here's a variation on that using the auto dataset:
sysuse auto
cap drop mpg_exc
gen mpg_excl = .
foreach n of numlist 1/5 {
qui regress mpg weight displacement turn if rep78 != `n'
qui predict hat
qui replace mpg_excl = hat if rep78 == `n'
qui drop hat
}
l mpg mpg_excl rep78
-----------------------------------
Thomas J. Steichen
[email protected]
-----------------------------------
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Nuno
Sent: Friday, July 06, 2007 4:08 PM
To: [email protected]
Subject: st: Running a regression for a subset of observations issue
Hello everyone,
Does anyone know if Stata is able to cyclically run n regressions on a n
observations sample, but at each ith cycle the ith observation is excluded
from the sample that will be used in the estimation?
The idea is the following: I have a sample of firms which are classified in
a particular SIC code. There is a model (Jones (1991)) that for each firm in
a given SIC estimates a regression based on the firms that compose that SIC,
excluding the firm being analyzed and then uses the estimated coefficients
to determine the expected value of a given variable for the excluded firm.
This might seem confuse, but the steps are quite straightforward (although I
can't implement it in Stata...).
So imagine we have in a given SIC code 10 companies. Then, the procedure
would do the following:
1. Run the regression Y=a0+a1*X1+a2*x2 and obtain the coefficients estimates
for all the 9 observations excluding the 1st one; 2. Use the estimated
coefficients to determine the expected value of Y for the first firm; 3.
Repeat 1 and 2 for the 2nd, 3rd, 4th ... firms in that SIC, excluding the
2nd, 3rd, 4th ... firm in each step.
Controlling for each SIC code is easy (just use -foreach-) but I'm unable to
implement steps 1-3...
Any ideas?
Best
Nuno
*
* 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/
-----------------------------------------
CONFIDENTIALITY NOTE: This e-mail message, including any attachment(s),
contains information that may be confidential, protected by the
attorney-client or other legal privileges, and/or proprietary non-public
information. If you are not an intended recipient of this message or an
authorized assistant to an intended recipient, please notify the sender by
replying to this message and then delete it from your system. Use,
dissemination, distribution, or reproduction of this message and/or any of
its attachments (if
any) by unintended recipients is not authorized and may be unlawful.
*
* 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/
*
* 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/
