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

Re: st: linear probability model (LPM)

From   Maarten buis <[email protected]>
To   [email protected]
Subject   Re: st: linear probability model (LPM)
Date   Mon, 14 Jan 2008 09:01:50 +0000 (GMT)

--- Tom Trikalinos <[email protected]> wrote:
> Under known circumstances odds ratios and risk ratios are numerically
> very similar ... all too often they are not... and then this
> misinterpretation bites... 

Now you are blaming the odds ratio for not being what it is not. The
whole idea of approximating risk ratios with odds ratios is unnecesary.
The odds ratio has a clear and understandable logic of itself. It is
this approximation that leads to the confusion. 

I do sympathise: if you go back in the statalist archive you will
probably find a post by me claiming that odds ratios are difficult. The
main reason for me was that the LPM so nicely fits with what I already
knew about regression (as was mentioned earlier, it can actually be
estimated using -regress-).

But you only need to understand two simple things to understand the
odds ratio: odds (number of successes for every failure) and ratios,
both of which have a clear substantive interpretation. When dealing
with categorical dependent variables, taking time to understand that is
a relatively small investment but a big payoff.

Now I can't resist to repeat my favourite feature request: I would like
Stata to output the baseline odds (-exp(_cons)-) when it presents the
odds ratios. It is not only nice to know that the odds of success is
e.g. twice as large for men than for women, but it is also informative
to know whether the odds of success for women is e.g. 0.01 success for
every failure or 1 success for every failure. 

-- 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

Sent from Yahoo! Mail - a smarter inbox

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index