|
Abstract:
|
A nested case-control data model provides a natural connection between failure time data and case-control data. The semiparametric analysis methods lead to 'standard' conditional logistic regression, but also apply to a large class of individually matched case-control sampling methods. However, this framework does not tell us what designs may be useful for the given study goals and data resources. In this talk, I will illustrate how the approach was used to solve three design and analysis problems: a two-stage design, absolute risk estimation when the number at risk is not known, and the appropriate analysis of studies with random digit dial controls. My focus will be on developing intuition about the estimators, general characteristics of successful designs, and how the methods easily extend to accommodate special situations.
|
