Dear Erica,
Space syntax has always suffered of a (very) poor use of statistics,
which is understandable in a field populated by people with such
diverse background. This might have influenced the use of a few
measures for the reasons you mentioned (poor correlations, etc.),
which are not necessarily problems of the measures per se, but
(probably) of the treatment given to them.
The other reason is that most of the software available now use
distance-based models, such as Angular-Segment-Analysis, which have a
different theoretical-and-methodological basis, one that does not
really focus on graph-structures but on a interpretation of how
individuals navigate and read the system.
However, measures of ringness are still some of the most important in
network science/graph theory. The clustering coefficient, for
instance, is a measure of ringness - reflecting the number of
triangles. There are versions for the number of squares (grid
coefficients) - which make more sense for cities. I have some comments
here:
http://eprints.ucl.ac.uk/2694/
I found later that both kinds of coefficient suffer of size effects
and it is often better to use the raw number of triangles or squares.
I have used them to create classifications of cities.
Ringness, in general, reflects the local interconnectivity between
spaces, lines or whatever a node represents. I found, for instance,
that the number of squares (rings of 3 steps) reflects grid
intensification and identifies the same "patches" that metric-based
measures from Angular Analysis identify.
Unfortunately, I am not sure if there is space syntax software
available that use/implement them. Perhaps, if you are working with
small systems, you could use other kind network/graph software, but
none comes to my mind at the moment.
Best Regards,
Lucas Figueiredo
On 10/08/2009, Erica Calogero <[log in to unmask]> wrote:
> Hi there,
>
> This is a question to the old timers and software developers as well as
> current practitioners (AKA everybody)....
>
> What happened to measures of ringyness that were used in early (SLoS)
> analyses? Justin de Syllas' 1981 thesis refers to relative ringyness or R.R.
> otherwise known as distributedness. When and why was this measure abandoned
> or not implemented in the software? Was it found to be unreliable, replaced
> by another measure that measured the same thing, not found to correlate
> statistically with any experimental observations? Or another more prosaic
> reason? I would love to know what happened to it, and also the use of A,B,C
> and D space classifications. So if any current practitioners or software
> developers are using or have implemented any of these measures in their
> software or practice I would love to know.
>
> Kind regards,
>
> Erica Calogero.
>
> EngD student.
>
--
Lucas Figueiredo
|