A Spatial–Temporal Model of House Prices in Northern Taiwan

This paper proposes a spatial–temporal model for analyzing the spatial–temporal distribution of house prices. The model consists of three components: a global mean function, a first-order autoregressive model, and a non-stationary spatial model. The global mean function captures the overall spatial trend, while the non-stationary model represents spatial dependence among districts. The autoregressive model accounts for the temporal correlation in house prices. The global mean function is expressed as a linear combination of basis functions, and the non-stationary model combines basis functions with stationary processes. Model parameters are estimated using a constrained least squares approach known as positive Lasso, which enables simultaneous parameter selection and estimation. The model is applied to a dataset of house prices over 23 months across 45 administrative districts in northern Taiwan, revealing non-stationary structures in the house price data.