A Bayesian space varying parameter model applied to estimating fertility schedules.

Area of Science:

• Biostatistics
• Spatial Statistics
• Demography

Background:

• Analyzing vital rates in small areas presents challenges due to data sparsity and spatial dependencies.
• Existing generalized linear models (GLMs) often assume fixed covariate effects, limiting their ability to capture spatial heterogeneity.

Purpose of the Study:

• To develop a spatial Bayesian generalized linear model (GLM) that allows covariate parameters to vary smoothly across space.
• To extend disease mapping methodologies to naturally incorporate space-covariate interactions.
• To apply the model to fertility curve estimation and study the diffusion of low fertility in Brazil.

Main Methods:

• A spatial Bayesian approach is employed, extending traditional GLMs.
• The model allows for smooth spatial variation in covariate effects.
• Coale's fertility model is used within the GLM framework for age-specific fertility rates.

Main Results:

• Simulations demonstrate the advantages of the proposed spatial GLM approach.
• The model effectively captures space-covariate interactions, providing richer insights than non-spatial models.
• Application to Brazilian census data reveals patterns in low fertility diffusion.

Conclusions:

• The proposed spatial GLM offers a flexible and powerful framework for analyzing vital rates in small areas.
• It enhances understanding of how covariate effects vary geographically.
• The methodology is valuable for demographic research, particularly in studying fertility trends and spatial diffusion patterns.