SIMPLE AND MULTIPLE CORRESPONDENCE ANALYSIS (CASE STUDY: CIGGARETTES ADVERTISEMENT DATA IN NATIONAL TELEVISION CHANNEL)

• Main Author: Annisa Riefina, Dana
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39143
• Institution: Institut Teknologi Bandung

Correspondence Analysis is one of multivariate analysis method for processing categorical data. Simple CA is a visualization of two-way contingency table and can be extend to multiple correspondence analysis (MCA) for more than two categorical variables. Matrix decomposition is used to reduce dimension so it obtains principal coordinate which can visualized the association between three or more categorical variables. There are two matrix decomposition that can be used in MCA, Singular-Value Decomposition (SVD) or Eigen-Value Decomposition (EVD). Indonesian Drug and Food Supervisory Agency (BPOM RI) have a specialized division to supervise tobacco products such as ciggarettes and its promotion activities. Ciggarette advertisements that aired in national television is being supervised and recorded each day based on regulations. Most of the variables are categorical data. There were 5.281 cigarette advertisements in January 2018. Using Pearson Chi-Square and Phi-Square statistics, this data has filled association assumption in CA method. Simple and Multiple CA has been done to categorical variables cigarette company, television channel, and advertisement duration. Correspondence map for simple and multiple CA has been obtained, because there were big observations, MCA gives small inertia. So, for this problem simple CA is more recommended. MCA with SVD and EVD both give correspondence map with different coordinate but the same variable configuration.