The case-cohort design has been put forward as an alternative design in epidemiological follow-up studies which is less expensive than a full-scale cohort study. In this design, all cases are assembled together with a randomly selected sub-cohort which is then used as a comparison group, one advantage being that the same sub-cohort may serve as a comparison group for a number of different case series. However, the fact that all case groups are compared with the same sub-cohort creates a correlation between estimated effects of exposures on the different outcomes. This asymptotic correlation is derived using martingale central limit theory by studying competing risks data samples using a case-cohort design. In a Monte-Carlo simulation study, we study effects on the correlation of factors such as number of cases and size of sub-cohort.
No takes yet. Share an insight, caveat, or question.
Poul Ejnar Sørensen (2000) studied this question.