Here is an example which comes pretty close, I think (not very simple
I have to admit):
sysuse auto, clear
reg price mpg
est store ols, title(Price)
heckman price mpg, select(foreign = mpg trunk weight length) twostep
est store ss2step, title(Price)
est store ss2step2, title(Car type)
heckman price mpg, select(foreign = mpg trunk weight length)
est store ssmle, title(Price)
est store ssmle2, title(Car Type)
estout ols ss2step ss2step2 ssmle ssmle2, style(smcl) ///
cells(b(star fmt(3)) se(par)) ///
label eq(1:1:2:1:2, .:.:3:.:., .:.:.:.:3, .:.:.:.:4) ///
keep(#1: #2: #3: #4:) eqlabels("", none) ///
varlabels(#3:_cons athrho #4:_cons lnsigma) ///
mgroups(OLS 2-step MLE, pattern(1 1 0 1 0))
One issue is that athrho and lnsigma are displayed for the MLE model,
instead of rho and lambda. There are three different approached to
change this (all are somewhat involved; ask me if you need details):
- use the transform() option
- include the number as scalars in e() (using -estadd-; not that some
of the numbers are already there) and then tabulate them in the bottom
of the table using the stats()
- get the e(b) (and e(V)) from the model, modify it, and repost using
-erepost- (see -ssc d erepost-)
(Note that the last approach could also be used to extract single
equations from the models, so that the complicated -equations()-
option could be avoided.)
One more thing: it seems from you example that the -vacant()- cells
suboption does not work if -unstack- is specified. I'll try to fix
that.
ben
On Nov 21, 2007 9:19 PM, Dimitriy V. Masterov <[email protected]> wrote:
> Here's an example of some nonsensical code:
>
> #delimit;
> use auto, clear;
> reg price mpg;
> est store ols, title(OLS);
> heckman price mpg, select(foreign = mpg trunk weight length) twostep;
> est store ss2step, title(2-Step);
> heckman price mpg, select(foreign = mpg trunk weight length);
> est store ssmle, title(MLE);
> estout ols ss2step ssmle using est.tex, style(tex) cells(b(star
> fmt(%12.3f) vacant(-)) se(par fmt(%12.3f) vacant(-))) unstack label
> replace collabels(, none)
> varlabels(_cons Constant ) starlevels(* 0.10 ** 0.05 *** 0.01)
> drop(athrho: lnsigma:);
>
>
> Here's the output:
> & OLS & 2-Step & &
> & MLE & \\
> & _ & Price & Car type &
> mills & Price & Car type \\
> Mileage (mpg) & -238.894***& -240.659** & -0.104** &
> & -272.494***& -0.145***\\
> & (53.077) & (119.333) & (0.052) &
> & (74.468) & (0.048) \\
> Trunk space (cu. ft.)& & & 0.016 &
> & & -0.039 \\
> & & & (0.065) &
> & & (0.062) \\
> Weight (lbs.) & & & -0.002** &
> & & -0.003** \\
> & & & (0.001) &
> & & (0.001) \\
> Length (in.) & & & -0.003 &
> & & -0.008 \\
> & & & (0.034) &
> & & (0.037) \\
> lambda & & & &
> 3886.502***& & \\
> & & & &
> (1485.568) & & \\
> Constant & 11253.061***& 9691.184***& 8.537** &
> & 11805.390***& 11.720** \\
> & (1170.813) & (3454.182) & (4.346) &
> & (1921.850) & (4.697) \\
>
> I would like lambda estimate from the mills collumn to be part of the
> price equation for the 2-Step, and for the MLE price column to have
> the lambda, rho and sigma estimates with their std. errors.
>
> *
> * 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/
