Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | rraciborski@stata.com |
To | statalist@hsphsun2.harvard.edu |
Subject | st: spmat idistance error |
Date | Mon, 30 Jan 2012 09:42:25 -0600 |
Ajita Atreya <ajita63@hotmail.com> is trying to create a banded spatial-weighting matrix using the user-written -spmat- command but -spmat- complains: > I have 24,000 observations on property prices. I am trying to create an > inverse distance matrix with truncation (200m) to use in spreg command. > I used the following code to create inverse distance matrix: > spmat idistance longitude latitude, id(id) dfunction(dhaversine) > vtruncate(1/.2) normalize(minmax) banded > The matrix was created once but when I re did it I started getting the error > as: Object not created-bands must be<=5516 > The model I am using is : > spreg g2sls lnprice ........, id(id) dlmat(matrixname) elmat(matrixname) We would need to see all the syntax that Ajita typed to diagnose the problem. Meanwhile, I can offer some general thoughts. Ajita typed . spmat idistance longitude latitude, id(id) ... This statement should have created an spmat object named "longitude"; my guess is that Ajita meant to type something similar to . spmat idistance myobj longitude latitude, id(id) ... ^^^^^^^ and -spmat- would have created an spmat object named "myobj". Whatever the case, the syntax for -spreg- would not work as in . spreg gs2sls ..., id(id) dlmat(matrixname) elmat(matrixname) options dlmat() and elmat() expect the name of an spmat object and from the code Ajita showed us, there is no spmat object named "matrixname". Re: banded matrices, Ajita should read section 5 of the spmat working paper, available at http://econweb.umd.edu/~prucha/Papers/WP_spmat_2011.pdf, too see whether the sorting trick described therein helps with banding the matrix. Finally, if Ajita has not done so, Ajita should also update -spmat- and -spreg- (and -spivreg-) to the latest version by typing . ssc install sppack, replace The updated code is slightly more memory efficient than the old version. --Rafal rraciborski@stata.com * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/