Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: -ivregress- bug or feature?

From   "Brian P. Poi" <>
Subject   Re: st: -ivregress- bug or feature?
Date   Thu, 21 Mar 2013 12:08:37 -0500

On 03/21/2013 10:54 AM, Hiroyuki Kawakatsu wrote:

-ivregress- does not appear to detect collinearity appropriately
when the -noconstant- option is specified and a (user generated)
constant is included in the instrument list. Is this a feature or
bug? (the error appears to be generated in the `hidden' command
_rmcoll* in -CheckCollin- of ivregress.ado)

/*code to illustrate feature/bug*/
set obs 10
set seed 1

gen x1 = rnormal()
gen x2 = rnormal()
gen y = x1 - x2 + rnormal()
gen z1 = 1
gen z2 = rnormal()
ivregress gmm y (x1 x2 = z1 z2), nocon

the last line results in the error:
note: z1 omitted because of collinearity
equation not identified; must have at least as many instruments not in
the regression as there are instrumented variables

not clear what z1 is collinear with... -gmm- works fine with

gmm (y - {b1}*x1 - {b2}*x2), one inst(z1 z2, nocons)


This is a bug in -ivregress-.  It is ignoring the 'noconstant' option when it checks for collinearity between the excluded instruments z1 and z2.  Because it's not respecting the 'noconstant' option, -ivregress- is complaining that z1 is perfectly correlated with the constant term.

We'll get this fixed in a future adofile update.

   -- Brian Poi

*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index