For homework 5 (QBA), question 4A, were you looking for the corrected odds ratio rather than the risk ratio?
Also, on slide 47 of the QBA lecture part 1, should the selection factor not be 3.2 rather than 0.31?
Can you provide any bread crumbs for how to even approach answering this question?
I don't see how to determine the RR or CI intervals referring to pg 42 of Orsini's text.
thank you
See selection bias section in Orsini's paper starting on page 38 and slide 50 of the slides (Part 1)
Yes that's a typo on slide 47. Thanks for catching.
For question 4A, assume the formula for OR applies to RR as well.
Thanks for clarifying, Francois.
Regarding using the same formula to correct for selection bias with OR and RR:
Can you also make clear whether we're just assuming this for the purposes of this HW set or if this is mathematically true?
Given the equation, it would seem that you can only correct for selection bias this way with the OR. For example:
corrected OR = observed OR/selection bias factor
corrected OR = (observed a *observed d/observed b *observed c) / (selection prob a*selection prob d/selection prob b*selection prob c)
This doesn't seem to translate directly to the RR given that the RR is (a/a+b)/(c/c+d). The selection probabilities don't appear to divide neatly the same way.
Thanks for clearing up my confusion.
I would think it does not hold mathematically for RR indeed and you should use this formula for the purpose of this HW only.
I think the main point is that this is how it works for OR. I am just not sure if that's a typo in the HW question (and answer) given Dr Mayeda and last year's lecturer reviewed it so please consider for the purpose of this homework:
either the formula applies to RR
either that's a typo in the HW and we meant OR instead of RR
Under both assumption, you would end up doing the same thing.
In the mean time, I am waiting for Dr Mayeda's answer.
I think the main point is that this is how it works for OR. I am just not sure if that's a typo in the HW question (and answer) given Dr Mayeda and last year's lecturer reviewed it so please consider for the purpose of this homework:
either the formula applies to RR
either that's a typo in the HW and we meant OR instead of RR
Under both assumption, you would end up doing the same thing.
In the mean time, I am waiting for Dr Mayeda's answer.
The reason why I don't want to say for sure that a similar formula doesn't apply for RR is that one may define a different selection bias factor when intent is to use the formula for RR. (although I don't know how that selection bias factor should be defined)