Statalist The Stata Listserver

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

Re: st: how does regress picks which vars to drop if N<X?

From   "Tom Boonen" <>
Subject   Re: st: how does regress picks which vars to drop if N<X?
Date   Mon, 12 Mar 2007 18:03:32 -0400

Thanks Austin Nichols,

I understand why the vars are dropped and i am not complaining that
Stata does this. My question (and sorry for not making this clearer)
is how Stata decides which variables are dropped.

The problem is clear in the case of collinear Xs (one the colinear
ones must be dropped), but if there are more Xs than obs, it is
unclear to me how Stata determines which vars are dropped and which to
keep. Or is there a rationale to proceed?

Thanks. Tom

On 3/12/07, Austin Nichols <> wrote:
Tom Boonen --
The problem is one of less than full rank, similar to a regression
including collinear RHS variables, so there is an infinite set of
possible solutions.
 help _rmcoll, help _rmdcoll
for a related utility that handles collinear RHS variables,
and note you cannot see the source code for these or -regress-.
For more on Stata's approach, see

On 3/12/07, Tom Boonen <> wrote:
> if i run regress with more covariates than observations, some
> covariates will be dropped. How does Stata decide which variables are
> dropped. I could not infer this from the source code.

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index