Fig. 5: Schematic representation of the model selection pipeline.
The pipeline is based on the approach proposed by ref. 40, but it extends it by also evaluating the Spatially Lagged X and the Manski models. The pipeline relies on the LM, the LR and the MI to identify an adequate spatial regression model for each regression. The pipeline should be applied by starting from the OLS model and only progressing to the next model if the test meets the significance condition indicated on the arrow. The preferred model (or models) is identified when it is not possible to advance further in the pipeline.
