Record Details
Modeling Density Dependence in the Presence of Measurement Error
ScholarsArchive at Oregon State University
| Field | Value |
|---|---|
| Title | Modeling Density Dependence in the Presence of Measurement Error |
| Names |
Payton, Quinn
(creator) Murtaugh, Paul A. (advisor) Lesser, Virginia M. (advisor) |
| Date Issued | 2015-04-17 (iso8601) |
| Note | Graduation date: 2015 |
| Abstract | Density dependence is an ecological concept concerning the mechanisms of change in the size of a population. The inability to census ecological populations confounds approaches to identify and quantify the level of density dependence. Statistical tests which ignore the presence of measurement error tend to result in misspeci fied type I error levels, rejecting the null hypothesis of density independence at a higher rate than the stated significance level. The purpose of this dissertation is to improve two tests for density dependence. Several flaws in a parametric bootstrapping test using the likelihood ratio statistic are explained. The necessity of several alterations to the parameterization of the density dependence model are demonstrated including the need for an explicit trending term to the model. A simulation study is employed to demonstrate that these alterations result in a correctly specified significance level. Additionally, a method is presented for accounting for the presence of measurement error with a nonparametric randomization test. Cholesky transformation matrices are used to correct for the dependence in observations that is introduced by the presence of measurement error. Simulations demonstrate that this correction results in adequately specified significance levels. Multiple observations at each time point are a theoretical approach to explicitly quantify the magnitude of measurement error. Both the adjusted parametric bootstrapping and nonparametric randomization tests are adapted to incorporate the use of multiple observations. A simulation study is used to show that the parametric test can lead to higher levels of statistical power in situations where the form of assumed density dependence is known. However, the nonparametric test is shown to be the more effective test in general application. |
| Genre | Thesis/Dissertation |
| Access Condition | http://creativecommons.org/licenses/by-nc-nd/3.0/us/ |
| Topic | Density-Dependence |
| Identifier | http://hdl.handle.net/1957/56214 |
