Related Experiment Videos
Estimation of the design effect in community intervention studies
R M Mickey1, G D Goodwin, M C Costanza
1Department of Medical Biostatistics, University of Vermont, Burlington 05405.
Statistics in Medicine
|January 1, 1991
Summary
Estimating mortality rate variance in community intervention studies is challenging. This study introduces a method using a design effect, found to increase with time, to improve variance estimation for cluster sampling.
Area of Science:
- Biostatistics
- Epidemiology
- Public Health
Background:
- Community intervention studies often lack replication, complicating variance estimation.
- Accurate variance estimation is crucial for assessing intervention effectiveness.
Purpose of the Study:
- To develop and validate a method for estimating mortality rate variance in cluster sampling with limited replication.
- To investigate the impact of time on variance estimation in community intervention studies.
Main Methods:
- Calculated design effect by comparing single-stage cluster sampling variance to simple random sampling variance.
- Utilized state-wide county data for age-adjusted mortality rates.
- Empirically applied the method to breast cancer mortality data.
Main Results:
- Design effects varied with the time period for death accumulation, increasing from 1.1 (1 year) to 3.5 (8 years).
- Observed consistent design effects across three states and nine years.
- Developed a time-dependent model that accurately represented the observed design effect relationship.
Conclusions:
- The proposed method provides a reliable approach for estimating mortality rate variance in cluster sampling.
- The time-dependent nature of the design effect must be considered in longitudinal community intervention studies.
- The developed model offers a valuable tool for biostatistical analysis in public health research.