Nonparametric regression estimation using multivariable truncated splines for binary response data.

Area of Science:

• Statistics
• Econometrics

Background:

• Traditional Truncated Spline estimators are for quantitative data, limiting their use with binary outcomes.
• Binary response variables are common in real-world applications, necessitating specialized regression models.

Purpose of the Study:

• Develop a multivariable Truncated Spline nonparametric regression estimator for binary response data.
• Address the need for models that capture changing variable relationships in specific sub-intervals for binary outcomes.

Main Methods:

• Proposed a novel multivariable Truncated Spline nonparametric regression estimator for binary data.
• Utilized the Akaike Information Criterion (AIC) for optimal knot point selection.
• Applied the estimator to real-world datasets concerning public health and socioeconomic indicators.

Main Results:

• The Truncated Spline nonparametric regression method demonstrated superior accuracy compared to binary logistic regression.
• The model effectively handles relationships with changing patterns across sub-intervals for binary responses.
• AIC provided an effective criterion for selecting optimal knot points in the model.

Conclusions:

• The developed Truncated Spline estimator is a valuable tool for analyzing binary response data with complex relationships.
• This method offers improved estimation accuracy over standard binary logistic regression.
• The approach is applicable to diverse fields requiring nonparametric analysis of binary outcomes.