A likelihood ratio test for a patterned covariance matrix in a multivariate growth-curve model

Area of Science:

• Statistics
• Biostatistics
• Multivariate Analysis

Background:

• Multivariate growth-curve models are essential for analyzing repeated measures data.
• Parameter estimation in these models typically relies on error matrices.
• The validity of covariance matrix assumptions significantly impacts model interpretation.

Purpose of the Study:

• To develop a parameter estimator for multivariate growth-curve models that is independent of the error matrix.
• To construct and discuss the distribution of a likelihood ratio test for patterned covariance matrices.
• To illustrate the application of the proposed method with a numerical example.

Main Methods:

• Utilizing a multivariate growth-curve model framework.
• Developing a parameter estimator conditional on the validity of a patterned covariance matrix assumption.
• Constructing a likelihood ratio test for the patterned covariance matrix.
• Analyzing a numerical example with two treatment groups and three repeated measures for three response variables.

Main Results:

• The parameter estimator is shown to be independent of the error matrix when the patterned covariance matrix assumption holds.
• A likelihood ratio test for the patterned covariance matrix is developed.
• The distribution of the likelihood ratio test is discussed.
• A practical application demonstrates the method's utility.

Conclusions:

• The proposed method simplifies parameter estimation in multivariate growth-curve models by leveraging patterned covariance matrices.
• The likelihood ratio test provides a statistical basis for validating these matrix assumptions.
• This approach offers a more robust and efficient analysis of repeated measures data in specific contexts.