Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: strange (?) behavior of -egen group- function


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: strange (?) behavior of -egen group- function
Date   Tue, 13 Nov 2007 11:41:08 -0000

On the face of it, Zurab has been bitten by an anomaly in -egen,
group()-. 

Given no varlist, or not given any varlist, the function does not 
complain. This is perhaps a small surprise, not least because this case
is not discussed in the documentation. 

There is another quite different take. 

What Zurab supplied was a varlist: it was just a limiting case, namely
an empty varlist (I prefer not to say "missing varlist", because that
conjures up all sorts of ideas about missing values, not the issue
here). If you 
look at the code with 

. viewsource _ggroup.ado

you'll see that the program works fine with an empty varlist. Apart from
any exclusions implied by -if- or -in-, if no varlist is supplied then
all 
observations end up in the same group. That seems logical to me. 
Whether it was intentional I can't say, but that's immaterial.
Unconscious
logic is just as valid as conscious logic. 

It's StataCorp's call, but in cases like this I wouldn't disturb
longstanding code. 

There is not much context here, but presumably this is happening 
unlooked-for in a program or do-file of Zurab's. If so, the easiest 
thing is trap the problem just upstream, say by

if trim("`varlist'") == "" { 
	* code handling this 
}
else egen <newvar> = group(`varlist') 

Alternatively, strip off the overhead of -egen- and program what 
you want directly. 

Note that there are plenty of precedents for empty varlists, just 
a difference on whether the user really means nothing or really 
means everything possible (e.g. -list-, -summarize-). 

Nick
n.j.cox@durham.ac.uk 


Zurab Sajaia


I just found out that -group- function of the -egen- family works with 
missing varlist as well, i.e. egen newvar=group() doesn't issue an error
but 
create newvar that equals to 1. I was using -capture- to catch the cases

with missing varlist but it never returned an error code ...

I guess I'll need to change my program but was wandering whether this is
a 
bug or normal behavior (then it might need to be mentioned in the help).

btw, I'm (still) using Stata 9.2 ...

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index