A COMBINED APPROACH OF MULTIPLE CORRESPONDENCE ANALYSIS AND HIERARCHICAL CLUSTER ANALYSIS FOR PROFILING RELATIONSHIPS AMONG CATEGORICAL RISK PREDICTORS: A BLUETONGUE CASE STUDY

Abstract

Bluetongue (BT) is a non-contagious virus in the Reoviridae family that infects both wild and domestic animals. It causes economic losses and reduces infected animals' production and reproduction. A total of 233 apparently healthy animals were screened for BT. Profiles of health condition of animals were identified using multiple correspondence analysis (MCA) and hierarchical cluster analysis (HCA), and the impact of the change in disease condition of animals was explored by examining the subjective evaluation of the impact of risk factors like (age, sex, season, species, and locality) with regard to BT disease providing an insight into a dataset through information visualization and it presents a useful application for visualizing associations amongst variable categories. The first two MCA dimensions retained up to 27% of the total inertia contained in the data. The positive BT results, summer, and old animals categories were loaded in the first dimension, while negative cases, Al-mounfia and winter categories were related to the second dimension. HCA identified three clusters. Cluster 1 was characterized by frequent and largely exclusive seronegative BT animals 91.67 % of animals in the cluster were seronegative, negative BTV category is the most important and related to cluster 1 with positive v-test=8.75. Cluster 3 can named a cluster of seropositive BT, up to 88% of cases were seropositive. We can conclude that seropositive BT is associated with summer and old age categories, whereas seronegative BT is associated with young age and winter categories, and thus MCA and HCA provide convenient and easy-to-interpret analytical tools for assessing categorical data relationships.

Keywords

• categorical data
• multiple correspondence
• inertia
• factominR
• hierarchical cluster