Hey Alireza,

the procedure of the first part of your question, if I understand it correctly, is possible to do in e.g. ArcGIS or R. You can construct so-called thiessen polygons (also known als Voronoi diagram) and around your points. Each polygon defines then an area of influence around its sample point, so that any location inside the polygon is closer to that point than any of the other sample points.

And I do not really understand your second part. Could you maybe specify it?

Best regards,

Jewgenij

thank you Jewgenij

I want to analyze the relationship between crime point events and proximity to banks with Geographically weighted regression model. I made a fishnet and i want to aggregate point layers into it then question is how to aggregate them?? I know I could do it with intersect as an containment approach but i searching other ways because there may be a relationship in border of tow cell so it could no be distinguished

second part is that i have the number of crime events in each cell of fishnet but how to make a standard criteria with that

I'm waiting for your answers

thanks

Okay, let put it this way. If you have ArcGIS with Spatial analyst extension you could directly use the Geographically Weighted Regression model incorporated in (http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//005p00000021000000).

Otherwise, if you have just two parameters (distance to banks and crime points) why don't using bivariate approaches like bayesian estimator or frequency ratio method. Frequency ratio is straightforward and can be implemented within each GIS environment. Calculate a euclidean distance raster for the bank locations. Classify the distance into certain classes and compare the prior distribution (map density) of the crime points with the density in a certain class (class_density).

class_density = number of crime points in class/ class_area

map_density = total number of crime points/ total area

likelihood = class_density/map_density

Likelihood higher than 1 will indicate that the crime number is above the average distribution in a certain distance class. Lower than 1 that they are below the average value and exactly one will point to the average distribution of crime points. The distance classes will most likely not match your fishnet (by the way, is it a regular one or did you used realistic boundaries?), but you could aggregate the raster values to the fishnet cells by common statistical measures using (MEAN, MEDIAN or Quantiles).

The procedure can be enhanced if you distinguish your banks by the size or clientele and generate a weighted layer for each bank class. Afterwards you could overlay them to get an enhanced crime picture.

The easiest way to see a relation among these parameters is however simple plot of crime numbers (here you could also specify which type, if possible) against the proximity to the bank (e.g. as a histogram or line plot) and if the relation shows linear pattern to run the simple linear regresion on it. Once you have got the beta of the regression you can recalculate the proximity map and aggregate to your fishnet as suggested above usind statistical measures.

Hope you can gain some ideas.

Best regards,

Jewgenij

Great

your ideas was brilliant

big thanks