Nonparametric regression estimation using multivariable truncated splines for binary response data

Abstract

In recent years, Truncated Spline estimators in nonparametric regression for quantitative data have gained significant attention. However, in practical applications, it is common to encounter situations where the response variable is qualitative (binary). As a result, Truncated Spline nonparametric regression models designed for quantitative data cannot be directly applied to binary response cases. Therefore, a method is needed that able to handle the relationship between variables whose patterns change at certain sub-intervals, where the response is binary. This article aims to develop a multivariable Truncated Spline nonparametric regression estimator specifically for binary response data. The proposed method is applied to analyze unmet need achievement status in East Java Province, Indonesia, and the percentage of the poor population in Indonesia. The findings indicate that the Truncated Spline nonparametric regression method provides more accurate estimates compared to binary logistic regression. Some of the highlights of the proposed method are: • This research develops a nonparametric truncated spline regression model tailored for binary response data analysis. • Using the Akaike Information Criterion (AIC) to select optimal knot points. • Evaluating the performance of the proposed model in comparing performance of nonparametric Truncated Spline model for binary response and binary logistic regression with the data real application.

Keywords

• Binary response data
• Truncated spline
• Nonparametric regression