NZTA-vehicle-crash-data/expl.R

# Exploration on data
# Export images 700x400

# Libraries
library(ggplot2)
library(dplyr)
library(RColorBrewer)
library(scales)

# Load data from Rdata file
load("crashdata.Rdata")

# Alternative way to load data
#dat <- read.csv("finaldata_201809.csv")

# Relevel crash severity
dat$CRASH_SEV <- factor(dat$CRASH_SEV, levels = c("N", "M", "S", "F"))

str(dat)

yeartab <- table(dat$CRASH_YEAR)
mpy <- sum(yeartab)/(18.75)

# Obviously, data missing from 2018
ggplot(dat) + geom_bar(aes(x=CRASH_YEAR), fill=NA, col="black") +
  geom_hline(yintercept = mpy, linetype="dashed") + theme_classic() +
  labs(x = "Year", y = "Number of crashes")
table(dat$CRASH_FIN_YEAR)

# Check that severity is given as F if and only if nonzero number of fatalities
table(dat$CRASH_SEV, dat$FATAL_COUNT)

# Look at number of vehicles involved
levels(dat$MULTI_VEH)
table(dat$MULTI_VEH, dat$CRASH_SEV)

# Open road vs vehicles
vehtab <- table(dat$MULTI_VEH, dat$URBAN, dat$CRASH_SEV == "F")
vehtab

# Graph crash severity
ggplot(dat) + geom_bar(aes(x=CRASH_SEV), fill=NA, col="black") +
  theme_classic() + labs(x = "Crash severity", y = "Number of crashes") +
  scale_y_continuous(labels = scales::comma)

knitr::kable(table(dat$MULTI_VEH, dat$CRASH_SEV))

# Not all openroad crashes are on state highways
table(dat$URBAN, dat$CRASH_SH_DESC)

# Motorcycle crash severity
table(dat$CRASH_SEV, dat$MOTOR_CYCLE > 0)

# Just look at fatal crashes on the open road on state highways
rurhwy <- dat %>% filter(URBAN == "Openroad", CRASH_SH_DESC == "Yes")
frurhwy <- rurhwy %>% filter(CRASH_SEV == "F")

str(frurhwy)

knitr::kable(table(rurhwy$MULTI_VEH, rurhwy$CRASH_SEV))

ggplot(rurhwy) + 
  geom_bar(aes(fill = MULTI_VEH != "Single vehicle", x = CRASH_SEV), position="dodge") +
  scale_fill_brewer("Vehicles involved", type="qual", palette = "Dark2",
                      labels = c("Single vehicle", "All other categories")) +
  scale_x_discrete("Severity", labels = c("Not", "Moderate", "Severe", "Fatal")) +
  theme_classic() + theme(legend.position = "right")

rurtype <- as.data.frame.matrix(table(rurhwy$CRASH_SEV, rurhwy$MULTI_VEH))
rurtypefrac <- rurtype / rowSums(rurtype)
rurtypefrac["Single vehicle"]

colSums(rurtype)/sum(rurtype)