Roger Newson has written a large series of programs, and
a series of papers explaining them. You should check out
the paper pointed to by -findit enduser-; the printed version
of that is in Stata Journal at SJ 3(3): 245-269 (2003).
Other than that, the references closest to what I
think you want are also in SJ, namely SJ 2(2): 202-222 (2002)
and SJ 3(2): 185-202 (2003). The first is also available
on the web somewhere as a series of OHP slides for a talk
based on it.
As it's my job, I ought to push the Stata Journal here,
as it contains most of the expository writing on Stata you
can find, other than the manuals and the FAQs and
material based directly on those.
Recently, I wrote -makematrix- (on SSC) as one implementation
of the idea of setting up a matrix and then populating
it with results. That's not designed to work directly
with -by:-, partly because I really didn't want to work
hard at reinventing -statsby-. (Also, I went on vacation,
and development stopped.) However, as you demonstrated,
-statsby- doesn't deliver, directly, on your problem.
However, a prior use of -separate- opens up some possibilities:
. separate mpg, by(foreign)
storage display value
variable name type format label variable label
----------------------------------------------------------------------
---------
mpg0 byte %8.0g mpg, foreign == Domestic
mpg1 byte %8.0g mpg, foreign == Foreign
. separate turn, by(foreign)
storage display value
variable name type format label variable label
----------------------------------------------------------------------
---------
turn0 byte %8.0g turn, foreign ==
Domestic
turn1 byte %8.0g turn, foreign == Foreign
. makematrix, from(r(mean) r(ub) r(lb)) listwise : ci mpg? turn?
mean ub lb
mpg0 19.826923 21.147465 18.506381
mpg1 24.772727 27.703963 21.841491
turn0 41.442308 42.546889 40.337726
turn1 35.409091 36.074633 34.743548
In other words you could have a
varlist
and a
byvar
then you could
foreach v of var <varlist> {
qui separate `v', by(<byvar>)
local newlist "`newlist' `r(varlist)'
}
makematrix, from(r(mean) r(ub) r(lb)) listwise: ci `newlist'
Other approaches are possible by virtue of -makematrix-
being able to put stuff in a named matrix, which can then
be combined with other matrices.
There will be more on -makematrix- in SJ 3(4): 2003.
Nick
[email protected]
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of
> Wallace, John
> Sent: 30 September 2003 18:36
> To: '[email protected]'
> Subject: st: RE: RE: RE: RE: ci, return lists, and statsby commands
>
>
> Thanks for the reply, Nick. I guess I have a more general
> question about
> how to compile the results of a series of tests and display
> them in a single
> summary. Can you point me to a reference that discusses
> building an array
> of values and then performing operations on the array?
> Your suggestion
> about doing -statsby- for each variable and then merging
> the results seems
> like a step in that direction, but rather than using a
> series of saved
> datasets can you put the results into memory somewhere,
> merge them, then
> save the final result? There may be no advantage to doing
> this, other than
> not having a bunch of intermediate variable.dta files to delete.
>
*
* 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/
