Re: {MEDSTATS} Effect Size

[Peter Flom]

Both the risk ratio and the odds ratio are more or less what they sound like.

If we let a = risk and disease
b = risk and no disease
c = no risk and disease
d = no risk, no disease

then RR = (a/(a+b))/(c/(c+d))

OR = ad/bc

Then we can do some multiplying in the first ....to get
a*(c+d)/(a+b)*c

= (ac + ad)/(ac + bc)

so, if a is small, they will be approximately equal.

HTH

Peter


-----Original Message-----
>From: Sarah <sarah....@gmail.com>
>Sent: Dec 16, 2009 10:17 AM
>To: MedStats <meds...@googlegroups.com>
>Subject: {MEDSTATS} Effect Size
>
>Hi all,
>
>I am aware this is a very basic question, but wanted some assistance
>with this. I calculating the Risk Difference, Risk Ratio, and Odds
>Ratio for a table of data where people have recieved a vaccine or
>placebo, and then whether they became infected with the disease or not
>(2x2 table). How do I interpret each of the RR, RD, and OR
>differently? At the minute all my interpretations say basically the
>same thing, just with the different values calculated. Basically, what
>is the difference between the risk ratio, and the odds ratio when
>interpretating the results?
>
>Also, does my table need to be in any specific order? I am aware I can
>use computer packages to work these out, but for this bit of work I
>need to do them by hand!
>
>Thanks,
>
>Sarah
>
>--
>To post a new thread to MedStats, send email to MedS...@googlegroups.com .
>MedStats' home page is http://groups.google.com/group/MedStats .
>Rules: http://groups.google.com/group/MedStats/web/medstats-rules


Peter L. Flom, PhD
Statistical Consultant
Website: http://www DOT statisticalanalysisconsulting DOT com/
Writing; http://www.associatedcontent.com/user/582880/peter_flom.html
Twitter: @peterflom

[Sarah]

Thanks

How would you interpret the results?
Would you just say for both 'using vaccine (or whatever) reduces the
risk by.....' or is there a different way of explaining each more
accurately?

On 16 Dec, 15:40, Peter Flom <peterflomconsult...@mindspring.com>
wrote:

> Website:http://wwwDOT statisticalanalysisconsulting DOT com/
> Writing;http://www.associatedcontent.com/user/582880/peter_flom.html
> Twitter:   @peterflom- Hide quoted text -
>
> - Show quoted text -

[Abhaya Indrayan]

I calculate RR only for prospective study that moves from antecedent to outcome, and OR only for retrospective study that moves from outcome to antecedent. Thus both should not be calculated from the same data unless some specific conditions are met.

Also, since OR is based on retrospective study, I interpret it as odds of antecedent instead of outcome. It is here that many may not agree. 

It is known that OR approximates RR when prevalence is low. Here too some caution is required. This prevalence is among the subjects actually studied. In a classical test-disease setup, diagnostic test is generally used when there is prior suspicion. In suspected cases, the prevalence of disease is rarely low. A diagnostic test is many times used as an adjunct to the clinical profile, and clinical profile may contain a lot of information about the presence or absence of disease.

~Abhaya Indrayan