Multivariate Statistical Analysis of Economic Shocks Perception in A Growing Economy (Case Study – Abia State)

A sample-based procedure for selecting an optimal subset of variables (shocks) for a cross-sectional survey is studied. Often, researchers are faced with the problem of selecting optimum subsets of variables in the midst of enormity of variables that will very best predict a model. Such inquiry traverse through biological taxonomy, medical diagnosis, marketing, personal behavior and attitude studies which either by necessity or design are categorical multivariate data and will transform to informative data which are dichotomous or qualitative. This study analyzed 17 economic shock variables across a sample of 750 households. Utilizing the expansion of 2, the study modeled the joint probability distribution of binary shock responses. The Kullback-Leibler Divergence statistic (2NI), based on the framework by 7, the hypothesis of shock independence was tested. Linear Discriminant Analysis (LDA) was subsequently used to determine misclassification probabilities and validate a parsimonious model.The findings reveal a significant departure from independence (2NI = 143.79, p < 0.01), indicating that shocks such as Herdsmen Vandalization (X15) and Crop Failure (X11) are highly covariant (ρ = 0.31). Although Death of Household Head (X1) had the highest individual prevalence (20%), the parsimonious model focusing on X1, X11, and X15 yielded an Apparent Error Rate (APER) of 0.13%, suggesting these three variables are nearly perfect predictors of household vulnerability. The study concludes that economic vulnerability in Abia State is driven by a