For the real estate industry, statistical models are helpful to assess the price of real estate in one particular housing area. However, the price of a real estate unit can be influenced by various factors, like the housing age, geographical location, and house age, which brings many difficulties for the process of price appraisal. Our client tried to refer to some papers in the journal to get help with the data modeling and prediction task, but the statistical models suggested are way more complicated to understand, especially for non-statisticians. Also, though these models may give relatively accurate predictions, they may lose some interpretability. Since it is also important for our client to know how other factors affect the price, compared to giving out good final price estimations, we decided to construct a model that is easy to interpret and gives relatively precise price estimations.
Our final choice for our client is a spline model, which treats variables differently according to their values, allowing a more flexible model fit. Specifically, we developed a general additive model (GAM) with B-splines for some covariates. It performed well in the prediction task, and its results are also easy to interpret, which make it perfectly suit our client’s need.