An introduction to nonparametric regression in demographic research

Area of Science:

• Demography
• Statistical Modeling
• Computational Social Science

Background:

• Traditional parametric regression models require strong a priori assumptions about functional forms.
• Reductions in computing costs have enabled the development of computationally intensive statistical methods.
• Demographic research can benefit from more flexible and accurate modeling techniques.

Purpose of the Study:

• To introduce nonparametric (NP) regression methods to demographers.
• To highlight the advantages of NP regressions over standard parametric approaches.
• To demonstrate the application of NP regressions using real-world demographic data.

Main Methods:

• Discussion of bias-variance tradeoffs in regression models.
• Introduction to localized regression techniques.
• Application of bivariate and multivariate NP regression techniques.

Main Results:

• NP regressions provide greater flexibility and accuracy in modeling relationships.
• These methods free empirical analysis from the constraints of mathematically convenient shapes.
• Illustrative examples are provided using data from the 1988 U.S. National Maternal and Infant Health Survey.

Conclusions:

• Nonparametric regressions represent a significant advancement for demographic analysis.
• These methods enhance the ability of researchers to model complex relationships accurately.
• Demographers are encouraged to adopt these powerful new computational tools.