spmat - inverse distance matrix involving two or more observations at some locations

[Chirag Kasbekar]

Hi,

I am trying to create a firm-to-firm inverse distance matrix and there are several firms at the same geographical coordinates--this is a historical dataset (mostly 19th C) and the coordinates are at the city or county level.

Years ago, I _successfully_ used this code to generate it:

spmat idistance invdistmatrix_sp x_long2 y_lat2, id(id) dfunction(dhaversine, miles)

Having to use this code to create some new measures, I run into this error: "Two or more observations have the same coordinates"

Please help! Have tried a few workarounds but nothing has worked so far. I would prefer not to use another software package if I can help it since I want the measures to be consistent across my studies, but if there's no choice, would appreciate any suggestions.

Thank you in advance!

Best wishes,
Chirag

[Andrew Musau]

spmat is from the Stata Journal (FAQ Advice #12).

"Two or more observations have the same coordinates"

The error message is self explanatory. Check for identifiers with the same coordinates.

bys id x_long2 y_lat2: gen tag=_N>1
list id x_long2 y_lat2 if tag, sepby(id)

ADDED IN EDIT: On first reading, I missed:

there are several firms at the same geographical coordinates--this is a historical dataset (mostly 19th C) and the coordinates are at the city or county level.

As I do not use these commands, I can only recommend trying out the official command spmatrix to see if you get the same error. If it does not make a difference, you can contract the dataset based on latitudes and longitudes, generate the measures and merge back with the full dataset.

[Chirag Kasbekar]

Hi Andrew,

Thanks very much for your response. I really appreciate it.

I did try spmatrix and run into the same problem. (A problem I didn't face many years ago!) I am not sure how to use contract and expand/merge here since what I need is an (symmetric) inverse distance matrix involving each firm. If I create the inverse distance matrix only with one firm from each location, I don't know how to include the other firm observations at those locations in the symmetric inverse distance matrix. The objective really is to create a spatially weighted competitor density (count) measure for each firm.

Thanks again!

Best wishes,
Chirag

[Andrew Musau]

I have looked into it and this replicates the error.

copy https://www.stata-press.com/data/r16/texas_merge.dta .
copy https://www.stata-press.com/data/r16/texas_merge_shp.dta .
use texas_merge, clear
*CREATE DUPLICATE
expand 2 if countyname=="Houston"
*THIS REPLICATES ERROR
spmatrix create contiguity W

The issue isn't duplicate coordinates per se, but duplicate identifiers. You need to assign each duplicate a distinct identifier and include the identifier in the shape file.

[Takuya Takahashi]

Hi, Chirag!

Have you solved your problem? Even though I am not familiar with STATA, I guess panel data analysis (spxtregress) would work for your analysis.