2

I am looking for either an answer or a resource that will outline the steps I need to take to perform a GWR, including screening variables and the interpretation.

I have access to ArcMap with all extensions, R, GWR4 and good old excel and pencil and paper. I have my data in a .shp file and in an excel file with the XY coordinates for the polygon centroids calculated.

I have been trying to piece this process together by reading research papers, text books etc, but aside from seeing a multitude of the same matrix equations and one video tutorial using ESRI which I am not that confident in, I am really lost on where and how to begin the actual analysis.

Note: If possible, I would like to avoid using the GWR algorithm in ESRI environment. I do have ArcMap for mapping if needed.

  • 1
    The best resource I could find was bisolutions.us/A-Brief-Introduction-to-Spatial-Regression.php. In general, if you are looking to understand an algorithm, I would avoid the "there is an ESRI extension" for this approach, as you will learn next to nothing about the underlying process from the black box approach. – John Powell Oct 14 '14 at 19:42
  • Could you give more information on what you are trying to calculate? – John Powell Oct 14 '14 at 19:48
  • Yes. I have rates of an outcome as my dependent variable for 270 polygons. I also have about 32 possible explanatory variables which may contribute to my dependent variable rate. I am trying to find which of these 32 best model (if at all) my dependent variable. – I Heart Beats Oct 14 '14 at 19:50
  • OK, sorry I asked :-) – John Powell Oct 14 '14 at 19:52
  • Actually, I have been working on some crime analysis recently involving 100,000 voronoi polygons and trying to weight by neighborhoods and then figure out if there is any spatial correlation. Doing it all with Postgres/Postgis queries, which is kind of fun. But 32 independent variables, different kettle of fish. – John Powell Oct 14 '14 at 19:57
1

In ArcGIS follow these steps.

  1. Put feature class in a file geodatabase.
  2. Drop feature class in map.
  3. Open attribute table to assure your fields are correct, make sure data types are correct and you do not have texts that should be long integers or vice versa.
  4. Visit the spatial statistics/modelling geographic relationships toolbox in arctoolbox.
  5. Select Geographically Weighted Regression
  6. Select your explanatory variable.
  7. Select dependent variables.
  8. Execute regression.
  9. Visit geoprocessing menu results for statistics such as R2, AIC, coefficients, etc.
  • I have done this already. I actually do not want to use ESRI unless I get a complete guide to interpreting the results. As, stated above, I am not confident in ESRI's GWR algorithm. But thank you though. – I Heart Beats Oct 14 '14 at 19:46

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.