Purity classification of the Çatalburun dog breed wth logistic regression analysis

This thesis aimed to determine the purity of Çatalburun dogs, a local hunting dog breed reared in Tarsus district of Mersin province, by using logistic regression models formed with various quantitative traits. In the study, 18 different morphological traits of 70 hunting dogs kept in Çatalburun Production and Rehabilitation Center affiliated to Mersin Metropolitan Municipality, and purity classification data scored by a specialist veterinarian were used. According to the results obtained with the examined univariate and multivariate models, it was determined that the performance of full model including all the independent variables was perfect with a very high AUC value of 0.98, and the ROC curve this model also showed a behavior very close to the ideal situation. The classification accuracy of the full model was also found to be approximately 92.9%. The full model was followed by the reduced model consisting of the variables of nose diameter, nose one wing length, bi-temporal width, bi-auricular length, chest circumference, hind paw wrist thickness, and ear width with an AUC value of 0.972 and an accuracy of 88.6%. Although the full model showed the best performance in terms of AUC and accuracy, it was concluded that the reduced model is also recommended for the purity determination of the Çatalburun dog breed, since it also gave lower AIC and BIC values in this study.