Higher Dimensional Tables
All of the general principles of testing and estimation presented for three-factor tables also apply when there are additional classification factors. The main difference in working with higher dimensional tables is that things become more complicated. First, there are many more ANOVA-type models to consider. For example, in a four-factor table, there are 113 ANOVA models that include all of the main effects. In five-factor tables, there are several thousand models to consider. Second, a great many of the models require iterative methods for obtaining maximum likelihood estimates. Finally, interpretation of higher dimensional models is more difficult.
KeywordsGraphical Model Conditional Independence Saturated Model Likelihood Ratio Test Statistic Closed Chain
Unable to display preview. Download preview PDF.