Publications

The summary of this report is: (1) The Kernal Density Estimator (KDE) model using log data provides the most conservative estimates; (2) The Empirical Tolerance Limit (ETL) model provides the least conservative estimates; (3) The results for the Karhunen-Loeve (K-L) and Normal Tolerance Limit (NTL) models lie in between the extremes; (4) The NTL results ended up being as credible as any of the other methods. This may be related to the fact that the data appeared to fit a lognormal distribution for higher values of {beta}; (5) The discrepancy between these methods appears to widen for higher values of {beta} and {gamma}; (6) The reasons for the extreme difference in the KDE results depending on whether one uses the raw data or the log of the data is not clear at this time; and (7) Which model will best suit our needs is not clear at this time.