This repository contains code and sample data for creating a generic risk map using rare-events logistic regression in R. I have built this repository as a means to share some of the code and procedures I used for analysis on a private project requiring discretion. The original premise of these analyses was the assessment of spatial risk factors for a rare wildlife disease however, in practice these examples could be applied to any geospatial analysis asking the question: how does X spatial variable impact the likelihood of Y spatial sample occuring (if Y is a rare event)? The code in this repository primarly demonstrates geospatial analyses often left to software like QGIS or ArcGIS in R. For example, code included shows:
- how to calculate the density of linear features within a polygon
- how to calculate distance-to variables between points in space and between points
- how to extract values within a polygon from a raster
- how to tabulate an intersection between polygon and raster layers
Please consult Process.md for the proper workflow.
In order to demonstrate the proper steps of spatial analysis I have chosen a random location, the state of California, but real spatial covariate layers to analyze: elevation, roads, and landcover. On principle, the "case data" provided does not represent any specific wildlife disease and has been produced entirely randomly for reproducible example purposes only (see sampledata.R).
This repository was built to serve as a part of my application to the Rstudio Summer Internship 2018 though the analyses for the original project began in September 2016.
All shapefiles were downloaded from publicly available sources: