The researchers developed a new method to estimate parameters in a flexible regression model. They used a spline smoothing approach instead of the commonly used kernel smoothing approach. Their new estimator outperformed other methods in a simulation study, showing the lowest mean squared error.