Hi!

I'm taking a class in applied categorical data analyses and I'm having a bit of trouble getting my head around what I think is a central part of the course.

With respect to sampling designs (A retrospective, prospective, case-control, cohort and cross-sectional), I'm not sure which measures of association (relative risk, odds ratio, and the risk difference) can be used.

I know this has a lot to due with how the data are collected. My Prof has sorta zipped through it, but this is all in the context of a 2 x 2 contingency table. He said something about how a study's design dictates what probabilities you can estimate (marginal, conditional, joint). That in turn dictates what measures you can use to test whether or not theres an association (relative risk, odds ratio, risk difference, likelihood ratio test statistic)

I'd really like some clarification on this or even better a web-page (I've tried searching but I haven't really turned up much, probably because of my search query).

I know my question is sorta wide open (and possibly vague), but I really need to organize these concepts in my head. Please bare with!


Cheers,

matt