101. Power and sample size requirements for non-inferiority in studies comparing two matched proportions where the events are correlated
- Author
-
Nam, Jun-mo
- Subjects
- *
SAMPLE size (Statistics) , *STATISTICAL matching , *COMPARATIVE studies , *STATISTICAL correlation , *CLUSTER analysis (Statistics) , *SET theory , *PARAMETER estimation , *STATISTICAL hypothesis testing - Abstract
Abstract: Consider clustered matched-pair studies for non-inferiority where clusters are independent but units in a cluster are correlated. An inexpensive new procedure and the expensive standard one are applied to each unit and outcomes are binary responses. Appropriate statistics testing non-inferiority of a new procedure have been developed recently by several investigators. In this paper, we investigate power and sample size requirement of the clustered matched-pair study for non-inferiority. Power of a test is related primarily to the number of clusters. The effect of a cluster size on power is secondary. The efficiency of a clustered matched-pair design is inversely related to the intra-class correlation coefficient within a cluster. We present an explicit formula for obtaining the number of clusters for a given cluster size and the cluster size for a given number of clusters for a specific power. We also provide alternative sample size calculations when available information regarding parameters are limited. The formulas can be useful in designing a clustered matched-pair study for non-inferiority. An example for determining sample size to establish non-inferiority for a clustered matched-pair study is illustrated. [Copyright &y& Elsevier]
- Published
- 2011
- Full Text
- View/download PDF