Monte Carlo Simulation of Uncertainties in Epidemiological Studies: An Example of False-Positive Findings Due to Misclassification

Alexander Shlyakhter and Richard Wilson

Proceed. ISUMA- NAFIPS '95 Third International Symposium on Uncertainty Modeling and Analysis, and Annual Conference of the North American Fuzzy Information Processing Society. IEEE Computing Society Press, 1995, pp. 685-689


The 95% confidence intervals for the Risk Ratios (RR) reported in epidemiological studies reflect only sampling errors and do not include uncertainty caused by misclassification and confounding. Analysis of uncertainties in epidemiological studies can be improved using Monte Carlo simulations. For case-control studies, we show how different misclassification of exposure status increases the probability of getting a statistically significant false positive result. The misclassification error is relatively more important when several studies are pooled together. Simulations enable the uncertainties in epidemiologic results to be reported similarly to natural science where systematic and statistical uncertainties are carefully combined. We illustrate this by showing how false positives can result from misclassification.

Keywords: Monte Carlo simulation, uncertainty analysis, epidemiological studies

Download: pdf
Back to papers