st: RE: RE: Odds ratio

["Visintainer, Paul" <Paul.Visintainer@baystatehealth.org>]

Rich is correct, as the references he provided show. I just want to note the popularity of the substitution method has much to do with the audience it appeals to. In their brief response to comment on their article, Zhang and Yu state, 

"The uncertainty surrounding the CI (ie, slightly narrower than it should be in some cases) might be indeed the trade-off between simplicity and precision.  Methods that can produce more precise estimates, yet are user-friendly, would be ideal." (JAMA, August 11, 1999; vol.282, p529.)

The "user" they refer to in "user-friendly" is likely to comprise those who see the CI as nothing more than a surrogate measure of the p-value.  Much of that audience -- many of them professional researchers -- would likely find this discussion quaintly academic, and would gladly trade a little "precision" for simplicity.

It's simplicity and ease of presenting it to some audiences are the reasons I have used it in the past.  I think we can expect this approach to be with us for quite some time.


-p

_______________________________________________
Paul F. Visintainer, PhD
Baystate Medical Center
Springfield, MA 01199

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of jhilbe@aol.com
Sent: Friday, April 09, 2010 12:44 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: RE: Odds ratio

I agree with Rich regarding the bias that can be produced in using 
oddsrisk. There
are some situations for which it is acceptable, but mostly it is not. I 
wrote
the command to provide it for those who wanted to use it - not to 
advocate its
use.  The method was proposed in an article by Zhang and Yu in 1998 in
JAMA -- Journal of the American Medical Association, presumably one of 
the
most elite of medical journals. So it gained considerable popularity.

I discuss the method in some length in my book "Logistic Regression 
Models",
pointing out its problems,but also when it can be used. I also offer 
general guidelines
of when it is acceptable to use a risk-probability interpretation for 
odds ratios, and
provide the calculations required to make the determination.

Joseph Hilbe




Date: Thu, 08 Apr 2010 16:21:49 -0400
From: Richard Goldstein <richgold@ix.netcom.com>
Subject: Re: st: RE: Odds ratio

well, first, I disagree with Paul's recommendation of -oddsrisk- there
is now lots of evidence that this algorithm is biased; for results, and
other ways of translating OR to RR see Blizzard, L and Hosmer, DW
(2006), "Parameter estimation and goodness-of-fit in log binomial
regression," _Biometrical Journal_, 48: 5-22; for a different way of
looking at this, with Stata code, see, Localio, AR, Margolis, DJ and
Berlin, JA (2007), "Relative risks and confidence intervals were easily
computed indirectly from multivariable logistic regression," _Journal of
Clinical Epidemiology_, 60: 874-882

second, if the OP really wants to present things in terms of changes in
probability, I recommend looking at Scott Long's -spostado- (-findit 
spostado-)

Rich

On 4/8/10 4:05 PM, Visintainer, Paul wrote:
> Rosie,
>
> Many health professionals find the ORs difficult to interpret, 
especially if
the base proportion is common.  It doesn't sound like the reviewer is
questioning your analysis, just the way you present the results -- and,
depending on your study, the ORs may look rather strange.  For example, 
if you
did a survey, and 60% of one group answered a question positively, 
while 90% of
another group answered it positively, your OR would be 6.0.  This can 
be quite
confusing for someone trying to interpret this OR like a relative risk.
>
> One approach might be to try a program by Joseph Hilbe, called 
-oddsrisk-.  It
will convert ORs from a logistic regression to a relative risk with 
95%CIs.  In
the example above, the OR was 6, but the conversion gave 1.5 (e.g., 
.90/.60 =
1.5).  This probably will make more sense to certain readers.
>
> Note that using the term "relative risk" depends on your study.  If 
your
logistic model was developed on a survey (e.g., cross-sectional), then 
your ORs
are prevalence ORs.  If you convert them using the -oddsrisk- program, 
you'll
have "prevalence ratios", not relative risks.
>
> Best,
>
> -p
>
> ________________________________________________
> Paul F. Visintainer, PhD
> Baystate Medical Center
> Division of Academic Affairs - 3rd Floor
> Springfield, MA 01199
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
[mailto:owner-statalist@hsphsun2.harvard.edu]
On Behalf Of Rosie Chen
> Sent: Thursday, April 08, 2010 2:48 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Odds ratio
>
> Hello, dear all,
>
> I have a question regarding a reviewer's comment on my use of odds 
ratio in
interpreting the results of a logistic regression, and would appreciate 
it very
much if you can provide any insight or any references for responding to 
the
comment.
>
> The reviewer commented that all results are expressed in terms of 
odds ratios
which makes it very difficult to assess the magnitude of the effect.
Probabilities and changes in probabilities would be much easier to 
interpret. My
impression is that, although it is true that predicted probabilities 
might be
easier to understand, odds ratios have been used extensively in 
research when we
interpret results from logit models.
> Do you have any suggestions regarding how to respond to this comment, 
or do
you have any statistics textbooks in your mind that recommend odds 
ratio as a
standard approach reporting results from logistic models?
>
> Thank you very much in advance!
>
> Rosie


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

----------------------------------------------------------------------
CONFIDENTIALITY NOTICE: This email communication and any attachments may contain confidential and privileged information for the use of the designated recipients named above. If you are not the intended recipient, you are hereby notified that you have received this communication in error and that any review, disclosure, dissemination, distribution or copying of it or its contents is prohibited. If you have received this communication in error, please reply to the sender immediately or by telephone at (413) 794-0000 and destroy all copies of this communication and any attachments. For further information regarding Baystate Health's privacy policy, please visit our Internet web site at http://www.baystatehealth.com.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/