[UrbanSim-Users] land price estimation and RPy

Hello,

We are trying to configure the land price model to estimate using RPy
so that we can use R's AIC and plotting features.

I see from urbansim.models.LandPriceModel.estimate() that the default
estimation procedure used is that defined in
opus_core.estimate_linear_regression.

I see that there exists another potential land price estimation
procedure in opus_core.estimate_linear_regression_r. In theory, we
would use an estimation procedure something like this one to take
advantage of R's AIC calculation and plot generation.

However, I don't understand how to configure the
urbansim.models.LandPriceModel to use something other than the
default estimation procedure.

Taking my cue from the documentation in
opus_core.RegressionModel.estimate(), I tried adding the key
'estimation' with the value 'opus_core.estimate_linear_regression_r'
to my estimation config. This doesn't work however, because
LandPriceModel (a subclass of RegressionModel) specifies a procedure,
and the logic in RegressionModel.estimate() only looks to the
estimation configuration for the estimation procedure if there is no
estimate procedure specified.

I temporarily--for testing purposes--modified RegressionModel.estimate
() to look first to the config. However the things I'm setting in my
estimation config (e.g. estimation, cache_directory) aren't in the
estimation config that RegressionModel.estimate() is seeing.

I'm unsure of what the best way is to do what I'm trying to do
(specify an estimation procedure other than that which is hard coded
in LandPriceModel.estimate()).

I may not be properly specifying the estimation procedure that I have
to run; how should this be done? (I didn't see any specific advise on
this in the Opus user guide.)

Or perhaps I should copy the LandPriceModel into a models directory
within my own package and modify that? (note: I modified
urbansim.models.LandPriceModel, as a test, to run
opus_core.estimate_linear_regression_r, and it did, only this
estimation procedure appears to be broken. However, this is a moot
point until I have an acceptable/"correct" way of specifying
estimation procedures to models).

Whether one of these or something else, what's the best way to proceed?

Thanks,

Brian Miles
Graduate Assistant
Spatial Analysis Lab - Rubenstein School of Environment and Natural
Resources
University of Vermont
Received on Wed Oct 03 2007 - 11:53:27 PDT