Related Experiment Videos
Estimating the degree of heterogeneity between event rates using likelihood
1Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, England.
American Journal of Epidemiology
|February 15, 1995
Summary
Epidemiologic studies often show variability due to random fluctuations. This study introduces a method to estimate true rate ratios, revealing significant heterogeneity in perinatal mortality rates.
Area of Science:
- Epidemiology
- Biostatistics
Background:
- Variability in mortality and morbidity rates in epidemiologic studies can stem from random fluctuations.
- This variability also affects rate ratios compared to reference rates.
Purpose of the Study:
- To apply a method for estimating the distribution of true rate ratios.
- To analyze perinatal mortality data from 515 small areas in England (1986-1990).
Main Methods:
- Combined Poisson variability with a gamma distribution assumption for true rate ratios.
- Utilized negative binomial log-likelihood for the gamma dispersion parameter.
- Employed direct numerical methods for maximum likelihood estimation and likelihood ratio tests.
Main Results:
- Empirical standardized mortality ratios (SMRs) ranged from 0 to 1.92 (5th to 95th percentile).
- Fitted gamma distribution indicated true rate ratio variability from 0.72 to 1.32 (5th to 95th percentile).
- Likelihood ratio test rejected the hypothesis of homogeneous true rates (p = 0.015).
Conclusions:
- The proposed statistical method effectively estimates true rate ratio distributions.
- Significant heterogeneity in perinatal mortality rates was detected.
- The method is computationally efficient and valuable for identifying areas needing further investigation.