Regressão linear em Componentes principais: uma aplicação em precificação de automóveis

Abstract

This work approaches an extensive database on cars in the United States, exploring their physical characteristics and sale prices. Through principal component analysis (PCA) and linear regression, we examine the influence of various factors on prices, aiming for insights into strategic decisions in the competitive automotive market. The methodology reveals the price elasticity concerning specific variables, facilitating informed choices. The study compares traditional linear regression models with those using indexes derived from PCA, highlighting the importance of dealing with multicollinearity. While the predictive capacity is similar, the PCA-based model avoids effect confusion caused by high correlations among covariates, emphasizing its stability in estimates. We report the applicability of linear regression in car pricing, crucial for maximizing financial results in a dynamic market. We do not provide a universal solution between methods choice but highlight the relevance of addressing multicollinearity in general cases.