We would like to calculate what will be the odds ratio difference between to models to show one model is significantly better than the other model.
We would like to 80% power and the case and control of the population will be 1000 caner and 4000 control. The variable we are testing will be divided into quartiles. The odds ration we are most interested is the highest quartile comparing to the lowest quartile.
I am very appreciated if anyone could share their ideas how to do it.
From your description it is hard to understand what it is you're going to test. Are the two groups significantly different? I think in order for somebody to help you, you'll need to clarify what are your null and alternative hypotheses.
Thank you for your reply. Here is what we are trying to do.
Null hypotheses: the new measurement to predict cancer is same as the conversational method
Alternative hypotheses: The new measurement has a significant higher odds ratio than the conversational method
we are going to recruit 1000 cancer subjects and 3000 controls and would like to calculate the power to detect the significant of higher odds ratio of the new method.
I have email SAS support and been told that at moment, proc power could not calculate this.
If you could think a way to do this that will be great help to us.
So there are two methods (measurements ?): new and conventional. There will be cases and control subjects, and the interest will be in predicting cancer risk.
Now imagine that you had already conducted your study and collected the data: what would be the model that you fit to your data in order to test the significance of the difference between the two methods?