commit 1fa8cb7393b1b831b4a976af1c613e85520f07ac
parent 43acea01bbe3b428bfe520a7b84f52cef9ea0ad7
Author: eamoncaddigan <eamon.caddigan@gmail.com>
Date: Thu, 21 Jan 2016 16:46:26 -0500
Mapping police violence + use of force project.
Diffstat:
| A | useOfForce.R | | | 68 | ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ |
1 file changed, 68 insertions(+), 0 deletions(-)
diff --git a/useOfForce.R b/useOfForce.R
@@ -0,0 +1,68 @@
+# Look at the relationship between departments' use of force policies and
+# violence.
+
+library(readxl)
+library(dplyr)
+library(tidyr)
+library(ggplot2)
+
+forcePolicy <- read_excel("UseofForcePolicyReview-bbdj.xlsx")
+# The column names are pretty long
+colnames(forcePolicy) <- sub(":.*", "", colnames(forcePolicy))
+
+# Strip answers down to TRUE (yes) / FALSE (no) values
+yesOrNo <- function(x) {
+ ifelse(grepl("^Yes", x),
+ TRUE,
+ ifelse(grepl("^No", x),
+ FALSE,
+ NA))
+}
+forcePolicy <- mutate_each(forcePolicy, funs(yesOrNo), 2:ncol(forcePolicy))
+
+# Now look at the mapping police violence project's report to find these
+# departments.
+mpvReport <- read_excel("MPVDatasetDownload-83zj.xlsx", sheet = 2)
+
+mpvReport <- mpvReport %>%
+ mutate(`Police Department` = sub("^[^[:alpha:]]*", "", `Police Department`),
+ row_number = 1:nrow(.)) %>%
+ filter(row_number <= 60) %>%
+ rename(`Police Department 2` = `Police Department`)
+
+# We'll do a fuzzy left join of these tables. This code is inelegant and
+# inefficient, but works OK for this few rows.
+forcePolicy[["Police Department"]][7] <- "Las Vegas"
+matches <- rep(NA, nrow(forcePolicy))
+for (r in seq_len(nrow(forcePolicy))) {
+ matches[r] <- agrep(forcePolicy[["Police Department"]][r],
+ mpvReport[["Police Department 2"]])
+}
+forcePolicy <- cbind(forcePolicy, mpvReport[matches, ])
+
+# Now let's see how these policies may effect police killings
+policyEffectiveness <- forcePolicy %>%
+ select(department = `Police Department`, 2:9, rate_of_killings = `Rate of Police Killings per Million Population`) %>%
+ gather("policy", "in_place", 2:9)
+
+# Plot it
+ggplot(policyEffectiveness, aes(x = in_place, y = rate_of_killings)) +
+ geom_point(aes(color = in_place),
+ position = position_jitter(width = 0.3, height = 0.0)) +
+ facet_wrap(~ policy) +
+ labs(title = "The effect of department policies on police killings",
+ x = "Policy in place",
+ y = "Rate of police killings (per million population)") +
+ theme(legend.position = "none")
+ggsave("all police force policies.png")
+
+# Wow. Zero in on de-escalation though.
+policyEffectiveness %>%
+ filter(policy == "Requires De-Escalation") %>%
+ ggplot(aes(x = in_place, y = rate_of_killings)) +
+ geom_boxplot(aes(fill = in_place)) +
+ labs(title="Policies requiring officers to de-escalate vs. police violence",
+ x = "Are officers required to de-escalate situations?",
+ y = "Rate of police killings (per million population)") +
+ theme(legend.position = "none")
+ggsave("de-escalation reduces police killings.png")
