The article explores different ways to estimate regression curves in spatial data. By comparing parametric, nonparametric, and semiparametric approaches, the researchers found that the nonparametric method produced the most accurate results. Specifically, using a nonparametric truncated spline approach led to the smallest error values, making it the most efficient model for spatial data with non-linear relationships between variables. This suggests that when dealing with spatial data, a nonparametric approach is better suited for capturing the complex relationships between predictor variables and responses.