Sheep,
One point of clarification on your bailing Lucas out - can you expand on : "
They are not small world graphs ( these require having a high cluster factor
that axial/convex maps cannot)" - can you demonstrate easily that axial and
convex maps CANNOT achieve high cluster factor? Or is this an empirical
statement about real urban systems - that we tend not to find such graphs in
reality?
All your other points read to me as though axial/convex maps are more
general than regular, planar, random etc. subsets of all possible graphs -
and so these for me would be the 'abnormal' (restricted subset out of all
possible graphs). This may just be semantics of course...
Alan
>
> To come to Lucas's rescue I would say syntactical ( or rather axial/
> convex) graphs are abnormal,
> They are not random
> They are not regular (latice,circle) graphs
> They are not planar graphs (common subset of graphs although building
> J-graphs can be)
>
> most importantly
>
> They are not small world graphs ( these require having a high cluster
> factor that axial/convex maps cannot)
> Urban graphs do have a non even distribution of connectives they tend
> to be closer to a Possion distribution than a exponential
> Urban Connectivity distributions do have a tendency to exhibit a
> double hump.
>
> I don't know if this is specfic to urban axial graphs but the
> correlation between total depth with in a radius R and number of
> items (nodes) within radius R correlates with r-squared of 0.999*
> (typically). Where R is less than the saturation (edge effect) radius.
>
> so yes I would say abnormal describes them.
>
> sheep
>
> On 2 Sep 2007, at 14:44, Lucas Figueiredo wrote:
>
> > Dear Fred,
> >
> > By j-graph I guess you interested in drawing 'normal graphs'. I would
> > recommend you to test these ones:
> >
> > NetDraw: http://www.analytictech.com/downloadnd.htm
> > Pajek: http://vlado.fmf.uni-lj.si/pub/networks/pajek/
> >
> > And my favourite: yEd
> > http://www.yworks.com/en/products_yed_about.htm
> >
> > They do not implement integration. However, you can replace
> > integration by closeness centraly as long you normalise the values
> > using the logarithm or the graph size. In other words:
> >
> > Integration ~ closeness / log(n)
> >
> > See Park 2005 for details:
> > http://www.spacesyntax.tudelft.nl/media/longpapers2/hoontaepark.pdf
> >
> > I am assuming that you do not need local integration as j-graphs are
> > usually explored in small systems.
> >
> > Good luck,
> > Lucas Figueiredo
> >
> > On 02/09/07, Frederico de Holanda <[log in to unmask]> wrote:
> >>
> >>
> >> Dear all:
> >>
> >> I forward the consultation I have made to Jorge:
> >>
> >> What are the software people are currently using to draw and measure
> >> j-graphs? Are they available through the internet? We have had
> >> problems in
> >> trying to download Jass, it has not worked so far. And New wave is
> >> totally
> >> out of date for unfriendliness... We have to have a more automatic
> >> procedure.
> >>
> >> Best and thanks.
> >>
> >> Fred
> >>
> >>
> >> Frederico de Holanda
|