|Date: ||Mon, 8 May 2006 14:52:28 -0400|
|Reply-To: ||Kevin Roland Viel <kviel@EMORY.EDU>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||Kevin Roland Viel <kviel@EMORY.EDU>|
|Subject: ||Re: help on power calculation|
|Content-Type: ||TEXT/PLAIN; charset=US-ASCII|
On Mon, 8 May 2006, wraong wrote:
> All I knew were:
> It's a case-control study on one certain exposure and a certain cancer. I
> don't know what the exposure is, I just know that it is a continous variable
> and will be categorized according to quartiles, compare the upper quartile
> with the lower one, the null hypothesis is OR=1.
First of all, you may be throwing away data. Why would you categorized?
What were the reasons provided to you?
Second of all, you will need additional parameters: N, effect size,
alpha, power, etc.
Is there any other measure of cancer available? Stage, severity,
biochemical assay, time to diagnosis, survival time, etc.? Why do you
want to perform a logistic regression?
What covariates are available? What is known about potential confounders?
What is the nature of the study?
Department of Epidemiology
Rollins School of Public Health
Atlanta, GA 30322