Related Experiment Videos
Analysis of correlated discrete observations: background, examples and solutions
Y H Schukken1, Y T Grohn, B McDermott
1Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, S1 072 Schurman Hall, Ithaca, NY 14853, USA. yhs2@cornell.edu
Preventive Veterinary Medicine
|July 2, 2003
Summary
This paper explains statistical models for correlated epidemiological data, including marginal, random effect, and transitional regression models. Understanding parameter interpretation across these models is crucial for accurate analysis.
Area of Science:
- Epidemiology
- Biostatistics
- Veterinary Epidemiology
Background:
- Correlated data is common in epidemiological studies.
- Traditional statistical methods may not adequately address data correlation.
- Accurate analysis requires methods accounting for dependencies within data.
Purpose of the Study:
- To highlight the use and interpretation of statistical techniques for correlated epidemiological data.
- To provide a conceptual background of statistical methods for correlated data.
- To emphasize the importance of selecting appropriate statistical models.
Main Methods:
- Overview of regression models for correlated data.
- Detailed explanation of marginal models, random effect models, and transitional regression models.
- Illustrative examples using veterinary literature data to demonstrate parameter estimation and interpretation.
Main Results:
- Statistical software for fitting correlated data models is increasingly available.
- Parameters in different model types have distinct interpretations.
- Examples illustrate practical application and interpretation of chosen models.
Conclusions:
- Choosing the correct statistical model is critical for valid epidemiological analysis.
- Understanding the nuances of parameter interpretation across different models is essential.
- The availability of advanced statistical tools necessitates careful model selection based on research objectives.