I have been trying to figure out how to achieve what I want with the help of your succinct advice. But I couldn't. The steps that I followed were:
1. First I researched what the "spatial weights matrix" was.
2. Then, I investigated which tools use it, and which of these tools can be used for my problem (distribute the amount of deposits among the population surrounding a branch).
The only tools that, I believe, can be used for my problem AND uses the "spatial weights matrix" was "Hot Spot Analysis (Getis-Ord Gi*)" (please, correct me if I am wrong). This tool can perform a hot spot analysis, that uses an spatial weight matrix.
3. Then, comes the task of creating the appropriate spatial weight matrix. As I want the distribution of bank deposits to follow the distribution of population, I thought that the most appropriate spatial conceptualization was to create a "cost distance" where the "cost" was the inverse of the population. This way, the deposits will spread farther in polygons more populated.
To built this "spatial weights matrix", I followed this steps:
a. Convert the Census block polygons to a raster, using "population" as the value.
b. Calculate the inverse of population for each cells, using the raster calculator. This was my "cost raster"
c. Calculate a "cost distance" using the previous raster.
Here is the result:
[ATTACH=CONFIG]19811[/ATTACH]
d. Here a get lost, as I don't know how to create the "input table" required by the "Generate Spatial Weight Matrix" tool.
Beside, this procedure is so convoluted that I have the feeling of being looking in the wrong direction.
Can you give me some advice?