Weight Object Question

[achapkowski]

I was reading the help and I was trying to figure out what some of the properties mean on the weight object.

1. What does s0, s1, s2 mean? I see the formula, but what do those number represent? The doc is lacking there. Could someone please shed some light on these properties?

[jGaboardi]

Also, the formulas aren't being rendered in the docs.

• pysal.lib.weights.W
• libpysal.weights.W

[sjsrey]

@achapkowski thanks for raising the issue. We will be fixing the rendering of the docs and will add more information about those measures. For now they are functions of the W matrix that are used by a number of tests for spatial autocorrelation and econometric tests/estimation methods.

[achapkowski]

@sjsrey thanks for the information, so it's a convenience property for other calculations?

Here is some doc on math in rst files: http://svn.python.org/projects/doctools/trunk/doc/ext/math.rst

[ljwolf]

@achapkowski Yes, they're fixed quantities useful to compute standard errors for spatial autocorrelation estimators and are there for convenience.

If you look at where they are in the code (e.g. Moran, Geary, Getis-Ord), you can see they have something to do with analytical expressions about the variance of random variables.

[achapkowski]

@ljwolf thanks for the follow up.

Is there some way to compare n number of weights (thinking of AIC or BIC) in pysal? So let's say I make 100 weights matrices off of the same dataset adjusting the input parameters accordingly, say a distance parameter, how can I assume one of better over another? Is there a metric for this?

That is why I started looking at the s0, s1, etc.. properties.

[sjsrey]

@ljwolf thanks for the follow up.

Is there some way to compare n number of weights (thinking of AIC or BIC) in pysal? So let's say I make 100 weights matrices off of the same dataset adjusting the input parameters accordingly, say a distance parameter, how can I assume one of better over another? Is there a metric for this?

That is why I started looking at the s0, s1, etc.. properties.

@article{anselin1986non, title={Non-Nested Tests on the Weight Structure in Spatial Autoregressive Models: Some Monte Carlo Results}, author={Anselin, Luc}, journal={Journal of Regional Science}, volume={26}, number={2}, pages={267--284}, year={1986}, publisher={Wiley Online Library} }

@Inbook{Florax1995, author="Florax, Raymond J. G. M. and Rey, Serge", editor="Anselin, Luc and Florax, Raymond J. G. M.", title="The Impacts of Misspecified Spatial Interaction in Linear Regression Models", bookTitle="New Directions in Spatial Econometrics", year="1995", publisher="Springer Berlin Heidelberg", address="Berlin, Heidelberg", pages="111--135", abstract="In most applied empirical studies in the realm of amongst others regional economics, urban planning and environmental economics, the estimation of ana priorispecified model is based on observations for a finite set of spatial units. This causes a number of persistent problems, which to varying extents have been dealt with in the literature. Essentially, the more fundamental problems boil down to the following issues.", isbn="978-3-642-79877-1", doi="10.1007/978-3-642-79877-1_5", url="https://doi.org/10.1007/978-3-642-79877-1_5" }

[github-actions[bot]]

Stale issue message