Estimation of a Semiparametric Varying-Coefficient Mixed Regressive Spatial Autoregressive Model

Area of Science:

• Spatial statistics
• Econometrics
• Public Health

Background:

• Understanding spatial dependencies in health outcomes is crucial.
• Existing models may not fully capture varying covariate effects in spatial data.
• Teen pregnancy rates are influenced by a complex interplay of individual, social, and environmental factors.

Purpose of the Study:

• To develop and validate a semiparametric varying-coefficient mixed regressive spatial autoregressive model.
• To investigate the impact of neighborhood and socioeconomic factors on teen pregnancy rates.
• To provide a robust statistical framework for analyzing spatially dependent health data with dynamic covariate effects.

Main Methods:

• Developed a semiparametric series-based least squares estimating procedure.
• Incorporated instrumental variables and series approximations for conditional expectations.
• Utilized a varying-coefficient mixed regressive spatial autoregressive model.

Main Results:

• The proposed estimators for nonparametric and parametric components are consistent.
• Asymptotic distributions of the estimators were derived and validated through simulations.
• The model effectively analyzes teen pregnancy data, highlighting neighborhood and socioeconomic influences.

Conclusions:

• The novel statistical model accurately captures complex spatial relationships and varying covariate effects.
• The method provides valuable insights into factors influencing teen pregnancy rates.
• This approach can be applied to other public health issues with spatial dependencies.