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: How to compare coefficients from ols and quantile regressions

From   Austin Nichols <>
Subject   Re: st: How to compare coefficients from ols and quantile regressions
Date   Tue, 27 Nov 2012 17:38:19 +0100

JVerkuilen (Gmail) <> and Buchs M. <>:

I think the poster is interested in comparing
dE(Y|X)/dX from OLS
dF_{.75}(Y|X)/dX from QR,
i.e. coefs, not individual predictions.

The trouble is the discontinuous nature of QR, and the lack of scores
and cluster-robust inference, which precludes use of -suest- and
related models.

I don't know about consistency of the bootstrap for comparisons across
effect on conditional means and conditional quantiles (calling
Stas...) but it is at least straightforward to implement:

sysuse auto, clear
g gpm=1/mpg
cap prog drop both
prog both, rclass
 reg gpm weight
 ret scalar ols=_b[weight]
 qreg gpm weight, q(75)
 ret scalar qr=_b[weight]
 su gpm if e(sample), mean
 assert r(N)==_N
 eret clear
bs ols=r(ols) qr=r(qr):both
test _b[ols]=_b[qr]

On Tue, Nov 27, 2012 at 5:21 PM, JVerkuilen (Gmail)
<> wrote:
> On Tue, Nov 27, 2012 at 9:54 AM, Buchs M. <> wrote:
>> Hi all,
>> I've run two regressions, one OLS and one quantile on the 75th percentile, using the same sample, dependent and independent variables (the dependent variable being household emissions). Now I'd like to test whether the coefficients for individual IVs from these two regressions are significantly different - is this possible and do you know how I could do this in STATA?>
> There's no particularly good reason that a QR for the third quartile
> would be directly comparable to the OLS coefficients that I know of,
> but perhaps I'm wrong. Medians often are, as long as the error
> distribution is symmetric and doesn't have heavy tails. What you might
> want to try doing is generate meaningful comparison cases and compute
> the predicted 75th percentile from both regressions.

*   For searches and help try:

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