The robustness of binary logistic regression and linear discriminant analysis for the classification and differentiation of BTV cases in goats
Azza B Musa, Amal Alsir Alkhidir Abedalraheem, H Hamad, Siddik Mohamed Ahmed Shaheen
Binary logistic regression (BLR) and linear discriminant analysis (LDA) are both applied in order to predict the probability of a specific categorical outcome based upon several explanatory variables (predictors).A random sample of 642 animals was selected from goats being represented by all predictors. The comparison between BLR and LDA was based on the significance of coefficients, sample size impact to percentage of correct classification rate, sensitivity, specificity and accuracy, and area under receiver operating characteristics (ROC) curve. Results showed both methods have similar contributors for data classification. The percentages of correct classification for total sample size were 87.3% for both models, and overall sample size the same trend was detectedin the percent of correct classification on both analyses as the sample sizes have been changed. The areas under ROC curve (AUCs) were 0.814 and 0.801 for BLR and LDA, respectively. However, BLR showed slight superiority for animals being correctly classified. In conclusion BLR and LDA can be used effectively for classification even with violation of normality assumption. The aim of this work is to evaluate the convergence of these two methods when they are applied in data from the epidemiological studies.