I have layer of point data of aid projects in Tanzania between 2000 and 2012, and a layer of point data for 42 villages where a survey was conducted in 2003, 2006,2009, and 2013. Within the survey point data my attribute table contains the percentage responses to a number of binary questions for respondents in each village. (There were 30 respondents from each village, so for example each village attribute table contains the % of people in x village who think corruption is a serious problem)

I want to see if the responses (dependent variable) are influenced by the proximity of villages to aid projects/the distance from aid projects (making proximity/distance my explanatory variable).

Should I perform OLS and then GWS? If so, how can I include distance as my explanatory variable? My data is currently geolocated in WGS_1984 datum, so should I change it to a projection?

  • A similar question was asked here.
    – Hornbydd
    Jul 9 '15 at 15:56

If I were doing this project, I would ignore GWS but do a bit of data prep before doing an OLS analysis.

Option 1: You can generate a "Distance to" field by performing a spatial join on your target layers. If you right-click the Villages point data in the Table of Contents and go to Join... you have the option to "Join data from another layer based on spatial location" in the first drop-down. Then select the Aid Projects points in the "Choose the layer to join to this layer" drop down. Next, the second radio option below states "Each point will be given...a distance field showing how close that point is (in the units of the target layer)." Note: this will probably give you some fields you don't need.

Option 2: If you have access to the Spatial Analyst extension, you could perform a Euclidean Distance operation on the Aid Projects point data, creating an outwardly-increasing measurement of distance from each point as a raster. Then, using Extract Values to Points on the Village point data using this Euclidean Distance raster as the input from which to pull, you'll get a field of distance to nearest Aid Project.

Once you have the distance field, it can be used as an explanatory variable in the OLS analysis.

GWS has other advantages, but in my mind this would be the most straightforward method.

Edit: Also, I would consider changing the projection to a more local system. You'll get a more accurate distance measurement. Consider UTM if your data fits within a single zone, or another more local system.

  • Thanks for the advice. I was using UTM for a projected coordinate systems but have some data that lies outside the 37s zone, would you have advice on a projected coordinate system that would be most appropriate for Tanzania?
    – Gina K
    Jul 17 '15 at 10:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.