Date of Award

3-20-2008

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Mathematics and Statistics

First Advisor

Samuel A. Wright, PhD

Abstract

The purpose of this research was to evaluate the appropriateness of using non-parametric estimators, specifically the Chao1, ACES, and Jackknife methods, for estimation of the number of unique species comprising a population. It goes on to develop a parametric method for the above stated problem. This research consisted of creating diverse populations, with known numbers of species, and applying the aforementioned non-parametric and parametric methods to samples drawn from the constructed populations. The parametric fitting of several different distributions to the sample data, including the lognormal, gamma, and Weibull was considered. Both types of methodologies were then applied to sample data from constructed wetlands, where little is known about the overall population size and species composition (number of unique species in the population). This research attempted to identify the underlying population distribution of the wetlands (via fitting of parametric curves to the sample), as well as focused upon demonstrating that the use of parametric methods were more apt to provide better results in estimating the number of species in a natural population. This research discovered the use of the non-parametric methods, developed originally for the use of smaller well-defined populations (Chao1) or computer debugging (ACES) was not appropriate for species estimation. The use of these methods resulted in lower bounds, which were several standard deviations away from the true number of species, for the contrived populations. This research found applying a parametric method was more accurate in representing the truth. A comparison of the two different approaches to species estimation and the advantages of using a parametric method over a non-parametric method are discussed as well.

AFIT Designator

AFIT-GAM-ENC-08-03

DTIC Accession Number

ADA481067

Share

COinS