This is an analysis of three groups’ likelihoods to classify crimes as gang-related. These results are still in preparation, so the analyses shown here use synthetic data, and do not reflect any real-world patterns.
I first model how the number of gang-characteristics relates to each rating group's likelihood of labeling a crime as gang-related. I also use these models to calculate each rating group's decision cut-points -- the number of characteristics for which their odds of classification pass the 50:50 mark. I then model how each characteristic independently relates to each rater's classifications.