Estimating uncertainty in physical measurements, observational and environmental studies: Lessons from trends in nuclear data

A. I. Shlyakhter, I. Shlyakhter, C. L. Broido and R. Wilson

Proceed. ISUMA '93 Second International Symposium on Uncertainty Modeling and Analysis, IEEE Computer Society Press (1993), pp. 310-317


The time trends in the sequential series of measurements of the same physical quantity show a general pattern: the normal distribution grossly underestimates the probability of large deviations from the true values; the probability is well described by exponential functions. By analogy with physical measurements, the results indicate that the usual 95% confidence intervals in epidemiology and environmental studies should be expanded to account for unsuspected systematic errors.

Keywords: uncertainty, confidence intervals, overconfidence, systematic errors, physical measurements, epidemiological studies, environmental studies

Download: pdf (864 K)
Back to papers