Personal Variable Relationship of Cancer Patients in Thailand Using Log-linear Models

Abstract

In this study, data sets of new cancer patients are used who were admitted
between January 2012 and December 2013 at the National Cancer Institute Thailand.
The main objective of this study is to test association between personal and cancer
related variables using log-linear models. A test of independence is used to find the
relationships between any two variables by chi-square tested and Cramer’s V. The
results in this study display that most paired variables of personal and cancer/
clinical variables are significantly related at p-value < 0.01. For both male and female
patients, the variable site of cancer is highly related to the stage of diagnosis, which
provides the highest Cramer’s V value. Moreover, the two-dimensional log-linear
models are used to find estimated parameters, expected frequencies and standardized
residuals obtaining the best model for the relationships between two defined
variables.