Setting up spatial regression

[Chen-Hao Tsai]

Hi all,

I have a dataset looks like the following structure:

FIPS	Project	Cost	var1, 2, 3, ...
10000	a	100
10000	b	120
10000	c	150
11000	d	110
11000	e	130

The FIPS is the county FIPS code. (I already have another shapefile in stata format with unique value in FIPS)

My goal is to investigate if project costs in FIPS 10000 is affected by project costs in FIPS 11000 (and other variables 1, ,2 ,3).

I went over the "spxtregress" however not sure if my data is a "panel spatial dataset."
First, I have multiple observations in one FIPS; second, the Project ID is not really a time variable, hence doing

xtset FIPS Project

is meaningless.

I appreciate any suggestions on how to set up the regression correctly.

Thank you.

[Mike Lacy]

(For the many non-U.S. participants here, I'd note that FIPS codes are U.S. government codes that designate local areas, such as counties, or smaller geographic units like census tracts.)

I only occasionally do spatial work, so I'm no expert, but you make no mention of spatial variables characterizing the location of your observations, such as latitude/longitude, or even information about the closeness of different observations to one another. Perhaps you have such data but have not mentioned it.

[Chen-Hao Tsai]

Mike Lacy I do have the spatial variables characterizing the location of my observations. For simplicity, those are var 1, 2, 3 ... in my simplified table above. Thanks.