Interval Estimation of Epidemiological Indices for Data Sampled from Clusters: A Review and an Extension

• Activity Number: 481
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Tasneem Zaihra*+ and Sudhir Paul
• Companies: University of New Brunswick and University of Windsor
• Keywords: Epidemiological Indices ; Clustered Binary Data ; Risk Difference ; Risk Ratio ; Relative Risk Difference ; Intraclass Correlation

Abstract

In epidemiology risk of developing a disease for individuals in an intervention group is compared with that in control group. The groups involve varying cluster sizes and binary responses within each cluster cannot be assumed independent. We discuss three epidemiological indices (EI) used to report the efficacy of treatment; Risk Difference RD, Risk Ratio RR and Relative Risk Difference RED. The preference of one EI over the other depends on design of study. Lui(2004) discusses several methods for interval estimation of these EI. In this paper we compare the methods discussed by Lui, using simulations, with the two new methodologies introduced in Paul and Zaihra(2008) for interval estimation of RR and RED. Simulations show that the method based on an estimate of variance of a ratio estimator performs best overall and it can be considered as an easily implementable alternate for all 3 EI.