Charlotte Housing Modeling Spatial predictors + Regression diagnostics
This project tests how neighborhood and amenity variables relate to housing prices using a regression workflow. The goal is not “just to run a model,” but to build a defensible specification, check assumptions, and interpret where the model performs well (and where it breaks down).
Build an interpretable model of housing prices using neighborhood-level predictors, while documenting the modeling logic and validating assumptions through diagnostic checks.
- Start with exploratory analysis to understand variable distributions and relationships.
- Use correlation screening and scatter plots to avoid unstable specifications and obvious redundancy.
- Fit an OLS regression model and evaluate performance using residual behavior rather than only fit metrics.
- Compare model variants and justify the “best” specification based on interpretability + diagnostics.
The project treats price as a spatial outcome and uses mapped predictors (income patterns, school presence, park density, and observed pricing patterns) to interpret results. Mapping is used as a diagnostic tool: to check whether relationships are geographically consistent and where local context may be driving model error.
The final model is evaluated through residual diagnostics and error analysis to identify where assumptions hold and where prediction/interpretation is less reliable. The outputs highlight which predictors align with price variation, and where spatial structure suggests missing variables or non-linear processes.
