INTERVAL BASED SAMPLE SIZE CALCULATION IN CASE-CONTROL STUDIES WITH MATCHED PAIRS UNDER INVERSE SAMPLING

Chii-Dean Lin (U. S. A.) and Kung-Jong Lui (U. S. A.)

Abstract

This paper provides a quantitative discussion on the effect resulting from the predetermined number of indexed matched pairs, defined as the pair in which the case is unexposed but the control is exposed, on the average length of the confidence interval. For a given ratio of the average length of the confidence interval relative to the underlying odds ratio, this paper provides a sample size calculation procedure to determine the required number of the indexed matched pairs under inverse sampling. To further facilitate use of the proposed sample size determination procedure and to help understanding of the effect due to the underlying mean and variation of the probability of exposure in the case group, this paper presents tables that summarize the required number of indexed matched pairs derived on the basis of the average length of the 95% confidence interval, the expected total number of matched pairs, and the minimum total number of matched pairs needed to obtain the required number of indexed matched pairs with a given probability in a variety of situations.

Keywords and phrases: sample size, case-control study, matched pairs, inverse sampling.