Regression Analysis of Real Estate Prices: A Practical Demonstration Emphasizing Different Objectives of Regression Analysis

Regression analysis is one of the most well-known methods in statistics. Although the method is relatively easy to understand and also easy to apply, in its common use the primary purpose for which the regression model is constructed is often neglected. This thesis aims to clearly present different approaches to regression analysis based on differing objectives. Specifically, these include descriptive, inferential, causal, and predictive analyses. Emphasis is placed on the specific assumptions, procedures, and interpretation of results of the individual analyses. While descriptive analysis does not require the fulfillment of any theoretical conditions, inferential analysis, by contrast, is based on several statistical assumptions related to the unsystematic component. The thesis further highlights the risks of causal interpretation and describes the difference between bivariate and partial relationships. The final analysis described is predictive analysis. The thesis focuses on point prediction, which is unique in that it is evaluated based on error on test data and is therefore not, unlike interval prediction, limited by assumptions about the unsystematic component. All analyses are demonstrated using the Ames Housing dataset. The software RStudio was used to perform all analyses.