How to handle discrete (count) data as dependent variable in Geographically Weighted Regression?

193
1
Jump to solution
09-08-2021 01:01 AM
Elijah
by
Occasional Contributor II

This may look a bit familiar but I am not sure about a few things.

I want to investigate the possible relationship between crime incidents and some environmental variables such as the:

Distance to roads, distance to residential areas, land use and population density. Note that the Dependent variable is  a count/discrete data (crime incident points) and all the explanatory variables are rasters and  continuous variables (land use is categorical, which I intend to include using the dummy variable route).

1. I intend to convert all the rasters including the land use raster to points. Is this really okay? 

2. Can I use the raw crime data (without aggregating to a polygon such as fishnet, hexagon, etc) in Geographically Weighted Regression (GWR) to assess the relationship as explained above. Or must I aggregate into some kind of polygon for it to work?

3. I was also looking at Forest-based classification... what do you think?

Kindly offer your thoughts, please.

 

 

0 Kudos
1 Solution

Accepted Solutions
DanPatterson
MVP Esteemed Contributor

How Geographically Weighted Regression (GWR) works—ArcGIS Pro | Documentation 

are you planning to use a Poisson (Count) Model type?

More suggestions and restrictions are listed in the link above


... sort of retired...

View solution in original post

0 Kudos
1 Reply
DanPatterson
MVP Esteemed Contributor

How Geographically Weighted Regression (GWR) works—ArcGIS Pro | Documentation 

are you planning to use a Poisson (Count) Model type?

More suggestions and restrictions are listed in the link above


... sort of retired...
0 Kudos