Date of Graduation
Master of Science in Mathematics
log linear models, categorical variables, odds ratio, contingency table, Wilks' statistic
In several social and biomedical investigations the collected data can be classified into several categories or groups. Often such data are represented by tables known as contingency tables. Statistical analysis of the contingency tables is done to examine the association among categorical variables. Log linear models are adopted to analyze higher dimensional contingency tables, and the association among the categorical variables is investigated. A suitable model is arrived at by fitting the various models such as the saturated model, homogeneous model, conditional independence models, joint independence models, and mutual independence model. The purpose is to find some type of independence among the variables, otherwise find the levels of association among the variables. Conclusions are drawn by analyzing a sample data. The sample data used illustrates that for high school graduates gender is jointly independent of race and family structure. For non-high school graduates no model for any type of independence fit well, and therefore, the levels of association between variables are calculated.
© Elena Antonia Castanada
Castanada, Elena Antonia, "Log Linear Models in Categorical Data Analysis" (2012). MSU Graduate Theses. 1646.