UNCERTAINTY
Data Uncertainty
Rule Uncertainty
o Crime events were generalized to the nearest map point by the Portland Police for anonymity (loss of precision)
o Modifiable Area Unit Problem: canopy cells are binary, therefore a cell is either canopy or non-canopy, which doesn't account for the variation in vegetation nor cells which may be a mix of canopy and non-canopy cover.
o Census data is based on 5 year estimates by the US Census Bureau, not actual counts which may mean a loss in accuracy.
o Some of the census data was anonymized by the census bureau because of small numbers of inhabitants.
o Kernel density estimation assumes continuous distribution of lights, crimes and population whereas there are pixels where the count will be 0.
o Population kernel density map was created from block group centroids but the population is usually distributed over the entire surface of the block groups.
o There are additional complex variables that will explain the distribution of crime which are not included in the regression models.