network weights

[knaaptime]

we're toying with the idea of adding a functiont to compute kernel weights on a street network using pandana

this doesn't exactly walk through the process of creating those weights, but shows how you could use pandana to get your network kernel

edit: not a kernel specifically this is a nice discussion of the problem

[jGaboardi]

pysal/spaghetti#258

[github-actions[bot]]

Stale issue message

[sjsrey]

KDE on networks is not as straightforward as in the 2-d planar space case. The complications are outlined in section 9.2.3 of

@book{okabe2012spatial, title={Spatial analysis along networks: statistical and computational methods}, author={Okabe, Atsuyuki and Sugihara, Kokichi}, year={2012}, publisher={John Wiley & Sons} }

These should be handled before we generate kernel based weights for the network.

[knaaptime]

i think we can work on kernel issues in a separate push. It still makes sense to have network weights even if they dont satisfy all the properties of a kernel

[jGaboardi]

Note that network weights are already possible with spaghetti.Network.contiguityweights (see code here). For an example see the Network Spatial Weights notebook.

[jGaboardi]

https://gist.github.com/sgillies/2217756