TABLE 4.

Spatial weight matrices comparison.

Spatial weight matrix	Pro	Con	Interpretation/Contribution
Method 1 - Distance	Simple to construct and understand Used often in literature	Less realistic	Convenient to use and easy to understand and interpret. Not realistic enough for real insight
		Inadequate for clustering
		Lacks temporal element
Method 2 - Mobile network	High spatial resolution Large amounts generated passively by mobile device users	Computationally expensive	Captures strong spatial associations over relatively short distances. Allows for the identification of patterns potentially missed by other methods
		Difficult to obtain
		Not representative
		Privacy concerns
Method 3 - Weighted Facebook data	Freely available data Potentially more representative	Low spatial resolution	Captures association between focal points of human activity regardless of distance
		Lacks specificity
Method 4 - Scaled Facebook data	Simple to construct and understand. Freely available data Potentially more representative	Lacks temporal elements	Adds additional information to previously simplistic model. Additional information improves clustering
		Low spatial resolution