Accession Number : ADA182565
Title : Minimum Distance Estimation of Mixture Proportions.
Descriptive Note : Master's thesis,
Corporate Author : AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Personal Author(s) : Benton-Santo,Robin N
PDF Url : ADA182565
Report Date : Dec 1986
Pagination or Media Count : 53
Abstract : Minimum Distance estimation was used to calculate estimates of the mixing proportion of the mixture of two normal distributions and the mixture of two exponential distributions. The estimation was carried out by using the Golden Search technique to minimize the Anderson-Darling goodness-of-fit statistic. A Monte Carlo simulation was run for both distribution mixtures, varying the mixing proportions from .25, .5 to .75 with sample sizes of 100 for the normal mixture and 750 for the mixture of exponentials. The simulation was run 500 times for each parameter combination. An ad hoc quasi-clustering technique was used to obtain the initial estimates for the parameters of the mixed normal while the method of moments technique was used to obtain initial estimates for the mixed exponential parameters. These estimates were then used to start the minimum distance routines which were used to obtain new estimates of the mixing proportions. Finally, the mean square errors were calculated for use as a means of comparison for the different estimation procedures.
Descriptors : *STATISTICAL TESTS, *STATISTICAL DISTRIBUTIONS, EXPONENTIAL FUNCTIONS, MIXING, MIXTURES, MOMENTS, MONTE CARLO METHOD, NORMAL DISTRIBUTION, RANGE(DISTANCE), SEARCHING, SIMULATION, THESES, ERRORS, ESTIMATES, PROBABILITY DISTRIBUTION FUNCTIONS, MEAN
Subject Categories : Statistics and Probability
Distribution Statement : APPROVED FOR PUBLIC RELEASE