0
$\begingroup$

Network generation parameters are to be optimised to produce spatial networks that match a given degree distribution (discrete: negative binomial) and a given continuous distance distribution (geographic distances to network neighbours: weibull). I need a possibly normalised score [0,1] to measure the distance between the empiric distributions of produced networks and given distribution functions.

$\endgroup$
  • $\begingroup$ Are you referring to a (network) graph? If so, how can you have a continuous distance? Normally, you would count the number of edges to transverse to get from one node to another. $\endgroup$ – gung Sep 22 '15 at 21:51
  • $\begingroup$ Thanks gung. I mean geographic distance in a spatial (social) network. I updated the question. $\endgroup$ – Sascha Sep 23 '15 at 8:16
0
$\begingroup$

I finally opted for the integral-based Hellinger distance since it can be easily scaled to [0,1] and is suitable for discrete and continuous distributions (also the R package distrEx provides an implementation). A valuable resource regarding the topic is a paper by Gibbs, A. & Su, F. 2002.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.