Page 1 |
Save page Remove page | Previous | 1 of 116 | Next |
|
small (250x250 max)
medium (500x500 max)
large ( > 500x500)
Full Resolution
All (PDF)
|
This page
All
Subset |
SIGNIFICANCE TESTINGS IN REGRESSION ANALYSES
by
Tan Hung Marie Ng
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
December 2008
Copyright 2008 Tan Hung Marie Ng
Object Description
| Title | Significance testings in regression analyses |
| Author | Ng, Tan Hung Marie |
| Author email | tanng@usc.edu; marieng@u.washington.edu |
| Degree | Doctor of Philosophy |
| Document type | Dissertation |
| Degree program | Psychology |
| School | College of Letters, Arts and Sciences |
| Date defended/completed | 2008-10-29 |
| Date submitted | 2008 |
| Restricted until | Unrestricted |
| Date published | 2008-11-11 |
| Advisor (committee chair) | Wilcox, Rand R. |
| Advisor (committee member) |
Baker, Laura A. Manis, Franklin R. McArdle, John J. James, Gareth |
| Abstract | Ordinary least squares is one of the most popular approaches for fitting regression models in the social sciences. However, it has been well documented that ordinary least squares is very sensitive to minor deviations from assumptions (e.g., Andrews, 1974; Dietz, 1987; Huber, 1973; Rousseeuw, 1984). Specifically, in the presence of outliers and heteroscedasticity, least squares estimates of coefficients can be misleading and can fail in capturing the genuine association among variables. Furthermore, when testing hypotheses regarding an individual regression coefficient, many have pointed out that the popular Student's t-test may have poor control over Type I errors and low statistical power under nonnormality and heteroscedasticity. In recent years, numerous new approaches for estimating and testing regression slopes have been introduced. However, certain properties of these methods are still not well understood. In this research, we looked more closely into the empirical performances of some of these new methods, specifically, their finite sample properties and resistance towards violation of assumptions. Moreover, we explored how some of these new methods can be applied to regression analyses in social sciences research, particularly for detecting interaction. Four simulation studies were carried out. The first study compared several test statistics which made use of the heteroscedastic consistent covariance estimators. The second study examined various methods for testing the slope of a simple regression model with a dummy variable. The third and fourth studies evaluated classic and alternative techniques for testing interaction effect. Our results suggested that although new methods generally perform better than classic approaches, there remain situations, in which no method performs satisfactorily. |
| Keyword | heteroscedasticity; interaction; nonnormality; outliers; regression; significance tests |
| Language | English |
| Part of collection | University of Southern California dissertations and theses |
| Publisher (of the original version) | University of Southern California |
| Place of publication (of the original version) | Los Angeles, California |
| Publisher (of the digital version) | University of Southern California. Libraries |
| Provenance | Electronically uploaded by the author |
| Type | texts |
| Legacy record ID | usctheses-m1762 |
| Rights | Ng, Tan Hung Marie |
| Repository name | Libraries, University of Southern California |
| Repository address | Los Angeles, California |
| Repository email | http://www.usc.edu/isd/libraries/services/ask_a_librarian/email/ |
| Filename | etd-Ng-2547 |
| Archival file | uscthesesreloadpub_Volume23/etd-Ng-2547.pdf |
Description
| Title | Page 1 |
| Full text | SIGNIFICANCE TESTINGS IN REGRESSION ANALYSES by Tan Hung Marie Ng A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (PSYCHOLOGY) December 2008 Copyright 2008 Tan Hung Marie Ng |
Comments
Post a Comment for Page 1
