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

Re: st: Binomial regression

From   Maarten buis <>
Subject   Re: st: Binomial regression
Date   Thu, 2 Aug 2007 22:14:37 +0100 (BST)

--- Constantine Daskalakis <> wrote:
> (ii) Causal effects interpretation (the RD can have it, but not the
> OR).

That surprises me. The thing about causal analysis is that it always
means an extrapolation outside your data. For instance the causal
effect  of gender consist of comparing the my probability of passing a
level of education with my probability of passing that level of
education if I were a female, the latter option is not observed and a
causal analysis consists of creating an extrapolation based on the
model of what whould have happend to my female alter ego. Such
extrapolations are very sensitive to any model misspecification, and
the identity link function seems to me very very dangerous, because it
fits the best fitting linear function inside the data, while at some
point the "real" function will have to taper off (otherwise it will
move outside the allowable range). So it may fit well to the data, but
is not suitable for extrapolation outside the data, which is a necesity
for causal analysis.
> (iv) Because the reviewer/editor/boss fancies it.

That is very discipline specific. In my discipline the identity link
function would not be accepted, or at least you would have a lot of
explaining to do (and would than still be rejected). 
> By the way, there is no inherently "more appropriate" link function
> for a particular type of outcome. It's just that some are technically
> easier than others.

True, there are a large number of link function from which a priori one
cannot say that one is inherently more appropriate than the other, but
some link functions inherently fall ouside that class of potentially
appropriate link functions, and the linear link has a very basic flaw
with respect to a binary dependent variable: it will produce
predictions outside the allowable range. 

-- Maarten

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Yahoo! Answers - Got a question? Someone out there knows the answer. Try it
*   For searches and help try:

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