So, I might have gone a bit over my head with the choice of this project. I feel I have a good grasp on the basics of regression, but I'm having trouble getting the output I desire in ArcMap.
I have a map of a city with census block polygons containing census data. On top of this map, I have 12 points located around the city, which contain "performance" attributes. Right now I'm just using an arbitrary 1 to 10 value as the performance value; I want to get the map to reflect the data how I want before I start my real exploratory regression analysis.
What I need the map to do is to find the explanatory variables within a certain radius of each of the points, and then use those results to predict performance (dependent variable) across the entire map.
I think my major issue is that not every block contains a point. Only 12 of 345 do. When I do a spatial join, the blocks that don't contain a point have a performance value of 0, rather than null, which wrecks the whole regression model. One idea I had was to create a 2 mile buffer around the points, intersect that buffer with the census blocks (so each census block that intersects with the buffer then inherits the performance value of the corresponding point), and perform Geographically Weighted Regression on that layer. The results are actually pretty good and the predicted values line up perfectly where the buffers overlap. Of course, this only calculates the prediction in the intersected buffer circles as that's all the layer contains, and not where the predictions would actually be of value.
Hopefully somebody understands what I'm trying to do and can provide some guidance. I attached an image of the map containing the census blocks containing the explanatory variables, the point locations containing the dependent variables, and the predicted results of the buffer+intersect+GWR.
