The background to my question is that I have unit sales for Product A aggregated by city. I also have unit sales for Category A aggregated by city.
The goal is to compare areas where Product A performs better than expected given performance of Category A. The reason for comparison is that unit counts tend to follow general population distributions (ie. at face value both data sets show same big cities are the primary markets).
What I did was take each data set and normalize it on z-score. I then computed two kernel density rasters, one for each set. Now my question is if there is a best way to compare the two given my objective?
Can I just use a raster calculation [Product A] - [Category A] to create a raster to show areas where the product outperforms the category? Is that a valid approach or is there a better way?
Also I noticed that the min/max value range computed for each kernel density is different for the two data sets. Does this matter for comparison?
- Product A: -0.0055 to +0.0253
- Category A: -0.0087 to +0.0683
Update The more I play with this, I'm not getting the results I need. I can't use division operators in the raster calculation because near-zero demonstrators are astronomically inflating the numbers. I believe I would have to re-normalize each of the two kernel densities first?