Some applications of the analysis of multivariate normal data with missing observations

Area of Science:

• Statistics
• Statistical Modeling
• Data Analysis

Background:

• Handling missing data in multivariate normal distributions is crucial for accurate analysis.
• Previous work established equivalence between maximum likelihood and generalized estimating equations for complete data.
• Generalized estimating equations were suggested for large sample sizes with missing data.

Purpose of the Study:

• To propose methods for analyzing multivariate normal data with missing observations.
• To evaluate the performance of generalized estimating equations compared to maximum likelihood.
• To formulate generalized linear models for diverse experimental designs.

Main Methods:

• Analysis of data from a multivariate normal distribution with missing completely at random observations.
• Application and comparison of generalized estimating equations and maximum likelihood methods.
• Formulation of generalized linear models for various experimental plans.

Main Results:

• Generalized estimating equations yield results similar to maximum likelihood under certain conditions.
• Equivalence holds when the covariance matrix is positive definite and smoothing is not required.
• The proposed methods are suitable for a broad range of experimental designs.

Conclusions:

• Generalized estimating equations are a viable alternative to maximum likelihood for analyzing multivariate normal data with missing values.
• The choice of method depends on sample size, missing data proportion, and covariance structure.
• The generalized linear models provide a flexible framework for complex experimental data.