Specification of the error covariance structure for linear mixed effects models with autoregressive characteristics: A simulation study
Item
-
Title
-
Specification of the error covariance structure for linear mixed effects models with autoregressive characteristics: A simulation study
-
Identifier
-
d_2009_2013:d5e97f3abf09:11181
-
identifier
-
11597
-
Creator
-
Jung, Jimmy,
-
Contributor
-
David Rindskopf
-
Date
-
2012
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Quantitative psychology | Educational psychology | Autocorrelation | Autoregressive | Error Covariance | HLM | Information Criteria | Linear Mixed Models
-
Abstract
-
This study examines the effects of specifying different error covariance structures on linear mixed models with autoregressive characteristics. Computer simulations were used to generate data varying magnitudes of autocorrelations, sample size, and series lengths. The data were fitted with error covariance structure specifications of compound symmetry, identity, autoregressive lag-1, Toeplitz, and unstructured. The effectiveness of using information criteria to correctly identify the error covariance structures was investigated and the impact of error covariance structure specification on estimates of fixed effects and tests of fixed effects were examined. In addition, a statistical power analysis of detecting the AR(1) autocorrelation parameter was conducted. Results provide recommendations on which information criteria to used for data with autoregressive characteristics, demonstrate how misspecifying the error covariance structure impact tests of fixed effects, and the data conditions necessary to accurately detect the AR(1) autocorrelation parameter.
-
Type
-
dissertation
-
Source
-
2009_2013.csv
-
degree
-
Ph.D.
-
Program
-
Educational Psychology
