Estimating relative risk from logistic regression probabilities?

[Irene P]

Hey there,

I was wondering if it's statistically "ok" to calculate the relative
risk comparing two groups by dividing their probabilities predicted by
the logistic regression model (e.g. men and women given the same values
on the rest of the predictors)? If so, how would I calculate a
confidence interval for this RR estimation?

The background to this question is that I would like to compare
different groups as to their probability to fall into the response
category. In the literature I found such comparisons made also by simply
calculating the differences between the probabilities of two groups of
interests. I was wondering if any one way is preferrable (ie. pr1/pr2 or
pr2-pr1)?

Any help or comments are greatly appreciated.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Albert-Jan Roskam]

Hi Irine,

The OR is one measure for relative risk, but it's not
necessarily the preferred one:
Alternatives for logistic regression in
cross-sectional studies: an empirical comparison of
models that directly estimate the prevalence ratio
Aluísio JD Barros and Vânia N Hirakata
BMC Medical Research Methodology 2003, 3:21

One assumption that has to be met is the rare disease
assumption, which states that the OR is only an
approximation of the RR when the prevalence is < about
10 %.

I don't understand why you would want to divide the
ORs - they are ratios already. If you include sex in
your model, and specify it as a categorical variable,
the OR is expressed relative to the other sex.

Oddly enough (no pun intended ;-) SPSS by default does
not give 95 % CIs when calculating ORs. You have to
indicate that.

The question whether you want to calculate a risk
difference or a risk ratio cannot be answered by
statistical arguments. It depends on your research Q.

Other people on this list know more about this than
me, but this was my 10 cents. ;-)

Cheers!
Albert-Jan


--- Irene Prix <[hidden email]> wrote:

> Hey there,
>
> I was wondering if it's statistically "ok" to
> calculate the relative
> risk comparing two groups by dividing their
> probabilities predicted by
> the logistic regression model (e.g. men and women
> given the same values
> on the rest of the predictors)? If so, how would I
> calculate a
> confidence interval for this RR estimation?
>
> The background to this question is that I would like
> to compare
> different groups as to their probability to fall
> into the response
> category. In the literature I found such comparisons
> made also by simply
> calculating the differences between the
> probabilities of two groups of
> interests. I was wondering if any one way is
> preferrable (ie. pr1/pr2 or
> pr2-pr1)?
>
> Any help or comments are greatly appreciated.
>
> =====================
> To manage your subscription to SPSSX-L, send a
> message to
> [hidden email] (not to SPSSX-L), with no
> body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send
> the command
> INFO REFCARD
>

Cheers!
Albert-Jan

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Did you know that 87.166253% of all statistics claim a precision of results that is not justified by the method employed? [HELMUT RICHTER]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


      ____________________________________________________________________________________
Check out the hottest 2008 models today at Yahoo! Autos.
http://autos.yahoo.com/new_cars.html

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Mark A Davenport MADAVENP]

All,

Does anyone know if any of the SPSS modules can do production frontier
analysis or if there are any other pieces of software that can perform
frontier analysis using SPSS or SAS datasets?



***************************************************************************************************************************************************************
Mark A. Davenport Ph.D.
Senior Research Analyst
Office of Institutional Research
The University of North Carolina at Greensboro
336.256.0395
[hidden email]

'An approximate answer to the right question is worth a good deal more
than an exact answer to an approximate question.' --a paraphrase of J. W.
Tukey (1962)

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Jason Burke]

Hi Mark,

Have you seen Frontier (Ver 4.1), developed by Tim Coelli. It is available from:

http://www.uq.edu.au/economics/cepa/frontier.htm

It only needs a routine to create the input files(s) for it. A trivial
task with the Cursor Calss in PYTHON, no doubt.

Cheers,



Jason


On 10/17/07, Mark A Davenport MADAVENP <[hidden email]> wrote:

> All,
>
> Does anyone know if any of the SPSS modules can do production frontier
> analysis or if there are any other pieces of software that can perform
> frontier analysis using SPSS or SAS datasets?
>
>
>
> ***************************************************************************************************************************************************************
> Mark A. Davenport Ph.D.
> Senior Research Analyst
> Office of Institutional Research
> The University of North Carolina at Greensboro
> 336.256.0395
> [hidden email]
>
> 'An approximate answer to the right question is worth a good deal more
> than an exact answer to an approximate question.' --a paraphrase of J. W.
> Tukey (1962)
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD