Re: st: Making vce(robust) the default

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 08:09 PM 8/5/2011, Stas Kolenikov wrote:
> On Thu, Aug 4, 2011 at 2:20 PM, Richard Williams
> <richardwilliams.ndu@gmail.com> wrote:
> > As for checking for the use of svy:, it can be done now in Stata 
> 12. Suppose
> > your program is called myprog.ado. Create another program called
> > myprog_svy_check.ado. svy: will check to see if this program exists, and if
> > it does it will execute it. For example,
> >
> > program myprog_svy_check
> >         version 12
> >         display "svy is being used"
> > end
> >
> > Maybe I can have it set some global parameter so the program knows svy: is
> > being used.
>
> That's a freaking dirty programming trick... or undocumented (and
> barely supported) implementation detail, depending on how you view
> this. It may be usable at the moment, but it may die off with the next
> release, or something. I mean, it's better than nothing, but you would
> want to program it very defensively, just for the case that something
> gets broken down: (1) your global is set by somebody else; (2) -svy-
> decides to call another program; (3) some other unpredictable
> circumstances that may cause wrong behavior of your program.

My approach is undocumented (in the manuals) but documented online. 
In Stata 12, type

help svy_parsing

For those not blessed with Stata 12 yet, part of the online help says

"Suppose that we developed an estimation command and added support 
for svy.  Let's assume our command is called myest and that there is 
an option called lrstats that is not allowed with the svy 
prefix.  When svy is used with myest, svy will look for a program 
named myest_svy_check.  If myest_svy_check exists, svy will call it 
with the options specified in the call to myest, along with the vce() 
option containing the VCE method that will be used."

The code that makes this all possible occurs early in 
_svy_check_cmdopts.ado, which is called by svy.

You are right that there is some risk. -help undocumented- says

"An undocumented command is a command of very limited interest, 
usually interesting only to Stata programmers, and is used by 
StataCorp in developing Stata. Undocumented commands are documented 
online, but their documentation does not appear in the printed 
manuals.  That undocumented commands are documented at all shows the 
"openness" of the Stata software. Undocumented commands may change 
their syntax or behavior in subsequent releases of Stata, though this 
is rare, so use them with caution."

I think this has the potential to be an extremely useful feature 
though, so I hope it sticks around for all eternity. ;-) My gologit2 
program sometimes works correctly with svy:, and other times it does 
not. I can fix that now with a gologit2_svy_check program.

I do agree that global variables can be problematic. I'll probably 
use more bizarre names. In my code I made a point of dropping the 
global as soon as it was no longer needed so it would be unlikely to 
stick around and haunt me.

If I had to find another way to do it - svy: uses iweights when it 
calls a program. So, you could check to see if iweights were being 
used. That would mostly create problems if by some wild chance the 
user actually did want to use iweights, so the approach I outlined 
above seems more foolproof to me.




-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/