Comparison of Uncertainty Calculation Models

By Hening Huang

Three models are available in the literature for calculating the expanded uncertainty using the experimental standard deviation: the Student’s t model, Craig model, and Bayesian model. This paper compares these three models by Monte Carlo simulation and by examining the random error and bias of the calculated expanded uncertainty. The results indicate that, among the three models, the Student’s t model is the least precise and accurate, the Craig model is more precise and accurate than the Student’s t model, and the Bayesian model is the most precise and accurate because of its use of prior information. When prior information is available, the Bayesian model is preferred for calculating the expanded uncertainty. When prior information is not available, the Craig model is preferred.

Read the Full Article (PDF)