I am preparing datasets for students in a spatial statistics course for a module on spatial autoregressive models. We will use ArcGIS 10.1. The first dataset will correspond to a mythical pre-industrial village in which houses closest to the tannery will be smaller and more tightly packed and as the distance from the tannery increases, houses will become larger and less tightly packed. Likewise, the property values of houses will increase as distance from tannery increases. NB: Tanneries were historically noxious points that smelled dreadfully and often polluted local waterways.

This example is purposely designed with multicollinearity between distance from tannery and size of house. I want my students to struggle with the question of what is explaining property values so they can understand what multicollinearity means in a concrete example. Later modules will include larger houses near the tannery, as well as smaller houses further from the tannery, so that my students can then use a spatial autoregressive model to identify the effects of positive externalities impacting a small house's property values when it located in a neighborhood with large houses, and the negative externalities impacting a large house located near a tannery.

I currently have a hand-drawn map of this mythical village, and have scanned it as jpg. Each building will have an ID number, a property value, a distance from the tannery and a size in square meters. While I could simply plug these data into Minitab, etc., I want to show my students how to use the spatial autoregressive model function in ArcGIS 10.1 (the proper name from the command escapes me as I type this, but it is under the Spatial Statistics Toolbar). For this function to work, however, I must also project my data. I plan on doing this by selecting a 300m x 300 m polygon somewhere on the globe and simply add control points that arbitrarily line up with my mythical village map (also scaled to capture a 300m x 300m area).

The added benefit of georeference this map is that I can also easily calculate the centers of these structures as well as their areas --- so long as I digitize the buildings.

Does this sound like a reasonable solution in terms of creating georeferenced and projected data to fit an image for an imaginary village?


2 Answers 2


Yes, your example seems logical, however I think you mean georeference not geocode. To start this setup you can do the following:

  1. Create a new polygon feature class defined with a spatial reference (in meters) that matches your geographic area of interest.
  2. Load in your polygon layer into ArcMap and start an edit session on it.
  3. Zoom in on the map to the area you want to create your 300mx300m polygon feature
  4. When clicking in your vertices I believe if you right click you should get some options for defining your segemnt lengths (e.g. 300m)
  5. Once finished editing, you can use the georeferencing toolbar to georeference your drawn image to the 300mX300m polygon layer.

I would suggest that before uploading the scanned map image to arcmap and georeferencing it, you "clean up" the scanned hand-drawn map in photoshop or other graphical editing program.

The preferred result should be a binary image with black features (roads, buildings etc,) and white or hollow background.

Because the spatial statistics tool require vector features, and the map is a raster image, you will need to convert from raster to vector. Doing so to a very messy raster image will result in bad polygon/line features that will need as much editing as drawing the map from scratch in arcmap.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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