[Openspace] Weight matrix in simulation
koschins at uiuc.edu
Thu Sep 29 12:04:34 CDT 2005
If I understand your question, you are asking if there is a way to
incorporate your knowledge of the data generating process in the structure
of the weights matrix. Your assumptions about how the spatial process works
come into play when you choose between different spatial weights matrices
that capture this process better or worse. For instance, with a first order
rook matrix, you are assuming that there are directional spatial effects
(e.g., North-South or East-West in a grid). The choice of a second order
queen matrix implies that the extent of your spatial effects includes all
immediate neighbors and the neighbors in the 2nd tier, so you have a broader
spatial effect than with a 1st order queen matrix. Inverse distance
weighting (which is not yet implemented in the current version of GeoDa)
assumes distance decay, and so on.
As far as GeoDa's weights are concerned, you are currently able to create
weights based on whether or not someone is someone else's neighbor (0,1).
You can incorporate these weights in simulations or modeling (free programs
such as R (e.g., the spdep package) allow you to convert GeoDa's weights to
other formats). Generalized weights beyond that will be incorporated in
future versions (no definite release dates are set yet).
From: openspace-bounces at sal.uiuc.edu [mailto:openspace-bounces at sal.uiuc.edu]
On Behalf Of Subhadip Chattopadhyay
Sent: Wednesday, September 28, 2005 11:49 PM
To: Openspace at sal.uiuc.edu
Subject: [Openspace] Weight matrix in simulation
I am confused about the structure of the weight matrix in simulation
exercises. The papers I have seen uses a weight matrix based on contiguity
in a regular lattice. The structure of spatial correlation is independent of
the nature of the variables. In empirical work researcher can exploit his
knowledge of the DGP to make an intelligent guess about the structure, is
there any procedure to emulate that in simulations?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openspace