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: rreg with robust standard error

From   Nick Cox <>
Subject   Re: st: rreg with robust standard error
Date   Tue, 17 May 2011 09:46:36 +0100

More positively, I would recommend -glm- which certainly supports
non-normal families, non-identity links and cluster structure.


On Tue, May 17, 2011 at 8:48 AM, Nick Cox <> wrote:
> I believe the short answer to be No.
> Expanding on this in various ways:
> 1. The help for -rreg- reveals no such option and the code for -rreg-
> reveals no such hidden option.
> 2. You seem to want in effect a hybrid of robust regression in which
> robust means protecting against non-normal errors, including possible
> outliers, and robust estimation in which robust means taking account
> of cluster structure. It's important to realise that these are
> different senses of "robust": although not so far as I can see
> absolutely incompatible in principle, they seem rarely brought
> together, even socially and not in software terms.
> 3. My guess is that it would be a substantial project to implement a
> hybrid and that anyone interested and competent would _not_ start with
> -rreg-, which is a relict from the 1980s. I think the precise theory
> would have to be created first, to say nothing of working out a
> software design and implementation. You may have references to the
> contrary (if so, please give them).
> If I've misunderstood your question, I imagine that you will need to
> clarify it. In particular, I don't know why you are both (a) claiming
> that something has been done and (b) unclear about whether it is
> possible.
> Nick
> On Tue, May 17, 2011 at 6:17 AM, lreine ycenna <> wrote:
>> I'm using the rreg command which produces standard error. However,
>> I've seen robust  (rreg) regressions being run with robust standard
>> error. I would like to know if it's possible to run rreg command and
>> have robust standard error in the regression?

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index