Regression Modeling Project - Real Estate Sales

This project involved developing a model to predict house prices based on various factors, including the transaction date, house age, distance to the nearest MTR station, number of convenience stores, latitude, and longitude.

​

The initial part of the project focused on exploratory data analysis to understand the potential significance of each variable. The price per unit of area was selected as the response variable. This was followed by fitting a simple additive model, which included all variables.

​

Subsequent steps involved refining the model by selecting only significant variables and checking for normality, linearity, and equal variance. When equal variance was not found, transformations were applied to the response variable to try to correct this. Despite these efforts, some challenges remained in satisfying all model assumptions.

​

The final model included five out of the six tested variables, providing useful insights for prospective homeowners. An unexpected finding was that longitude was not a significant variable, while house age was. The project highlighted the challenges and complexities of creating a robust regression model, including dealing with residuals and handling outliers.