Fitting a three-parameter lognormal distribution with applications to hydrogeochemical data from the National Uranium Resource Evaluation Program

Technical Report · Mon Oct 01 04:00:00 EDT 1979

DOI:https://doi.org/10.2172/5702331· OSTI ID:5702331

Kane, V. E.

The standard maximum likelihood and moment estimation procedures are shown to have some undesirable characteristics for estimating the parameters in a three-parameter lognormal distribution. A class of goodness-of-fit estimators is found which provides a useful alternative to the standard methods. The class of goodness-of-fit tests considered include the Shapiro-Wilk and Shapiro-Francia tests which reduce to a weighted linear combination of the order statistics that can be maximized in estimation problems. The weighted-order statistic estimators are compared to the standard procedures in Monte Carlo simulations. Bias and robustness of the procedures are examined and example data sets analyzed including geochemical data from the National Uranium Resource Evaluation Program.

• Research Organization: Oak Ridge National Lab., TN (USA)
• DOE Contract Number: EY-76-C-13-1664
• OSTI ID: 5702331
• Report Number(s): GJBX-175(79); K/UR-27
• Country of Publication: United States
• Language: English

Related Subjects

050200 -- Nuclear Fuels-- Exploration-- (-1987)
11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS
99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
DATA ANALYSIS
GEOCHEMICAL SURVEYS
GEOLOGIC DEPOSITS
MATHEMATICS
MAXIMUM-LIKELIHOOD FIT
MOMENTS METHOD
MONTE CARLO METHOD
NORTH AMERICA
NUMERICAL SOLUTION
PROSPECTING
RESOURCES
SIMULATION
STATISTICS
URANIUM DEPOSITS
USA