Statistical Analysis of Car Data Using Analysis of Covariance (ANCOVA)

Abstract

Engineering statistics involves data concerning manufacturing processes such as: component dimensions, tolerances, type of material, and fabrication process control. In this work, some statistical analysis was done on a car data which was collected in 1983. This data set contains a sample of 66 cars from three different countries producers (USA, Japan, and Europe) in terms of their price and 13 different features. The objective was to determine how these quantitative variables affect the car price using the Analysis of Covariance (ANCOVA). Some analysis was done to find the best Multiple Linear Regression (MLR) model that identifies the most significant factors that influence the car price. The addition of removal of new factors was based on trial and error and based on the value of the linear correlation with the car price (Pearson coefficient of correlation r). 9 out of 13 factors were chosen based on MLR and the model selection/reduction analysis. The conclusion from ANCOVA analysis found that the European cars are the most expensive cars when compared to USA and Japan. Japan cars are the second expensive and USA cars are the cheapest.