Spatial analysis demonstrates that controlling for confounding variables alters identified movement patterns in urban taxi trip data, indicating the true drivers of spatial interactions.
Key Points
To develop and assess spatial analysis methods that identify flow autocorrelation patterns while controlling for multi-sourced confounding effects.
Developed Partial Conditional FlowLISA (PC-FlowLISA) and Full Conditional FlowLISA (FC-FlowLISA) frameworks to account for confounding variables in spatial flow data.
Evaluated the proposed methods using publicly available taxi trip records from New York City.
Controlling for confounding effects significantly altered both the identification and geographic interpretation of spatial flow patterns.
The conditional approaches successfully distinguished true spatial interaction mechanisms from spurious associations induced by confounding variables.