Analysis of long repeated measures designs.
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1983.
|
| Subjects: | |
| Online Access: | Link to OAKTrust Copy Link to ProQuest Copy |
| Abstract: | Standard univariate procedures for repeated measures ANOVA are modified for the case where errors are hypothesized to arise from a covariance stationary process. Limits of (epsilon), the degrees of freedom reduction factor, are studied as the number of repeated observations increases. An estimator for (epsilon) based on the assumption of an autoregressive error process is proposed, and is shown to result in tests that compare favorably with those utilizing the true value of (epsilon). A generalized least squares model for repeated measures is also investigated, and the asymptotic distribution of a test statistic for testing a wide variety of hypotheses is derived for that model. A simulation study, however, showed that the statistic did not achieve its theoretical asymptotic distribution for many of the models studied.Several tests for detecting differences between error processes underlying different observations in the model are considered, all utilizing autoregressive models. Simulation studies showed that, unless the true order of the autoregressive process was known, the distribution of the statistics is difficult to model. |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Statistics." |
| Physical Description: | viii, 164 leaves ; 29 cm |
| Bibliography: | Includes bibliographical references (leaves 91-93). |