Abstract
A parsimonious covariance structure of repeated measures is often sought for purposes of increased power for testing hypotheses about the means, and for insight into the stochastic processes governing the repeated measures. For normal data, model selection is often based upon likelihood ratio tests or information criteria derived from the likelihood, sometimes supplemented with graphical plots of correlations and partial correlations. We exploit the ordered nature of repeated measures to decompose the likelihood ratio goodness-of-fit test statistic, and display graphical fingerprints associated with the covariance structures to help detect covariance structure misspecification, in order to provide guidance in choosing an appropriate structure for the data. The proposed methodology is illustrated with simulated repeated measures data and then applied to an experiment to compare tillage methods of pasture establishment.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Moser, E. Barry and Macchiavelli, Raul E.
(2002).
"MODEL SELECTION TECHNIQUES FOR REPEATED MEASURES COVARIANCE STRUCTURES,"
Conference on Applied Statistics in Agriculture.
https://doi.org/10.4148/2475-7772.1197
MODEL SELECTION TECHNIQUES FOR REPEATED MEASURES COVARIANCE STRUCTURES
A parsimonious covariance structure of repeated measures is often sought for purposes of increased power for testing hypotheses about the means, and for insight into the stochastic processes governing the repeated measures. For normal data, model selection is often based upon likelihood ratio tests or information criteria derived from the likelihood, sometimes supplemented with graphical plots of correlations and partial correlations. We exploit the ordered nature of repeated measures to decompose the likelihood ratio goodness-of-fit test statistic, and display graphical fingerprints associated with the covariance structures to help detect covariance structure misspecification, in order to provide guidance in choosing an appropriate structure for the data. The proposed methodology is illustrated with simulated repeated measures data and then applied to an experiment to compare tillage methods of pasture establishment.
