Setup

Load packages

library(ggplot2)
library(tidyr)
library(gridExtra)
library(rstanarm)
library(brms)
library(bayesplot)
theme_set(bayesplot::theme_default(base_family = "sans", base_size = 16))
library(patchwork)
library(loo)
library(rprojroot)
root<-has_file(".BDA_R_demos_root")$make_fix_file()

1 Introduction

This notebook demonstrates time series analysis for traffic deaths per year in Finland. Currently when the the number of traffic deaths during previous year are reported, the press release claims that the the traffic safety in Finland has improved or worsened depending whether the number is smaller or larger than the year before. Time series analysis can be used to separate random fluctuation from the slowly changing traffic safety.

2 Data

Read the data (there would data for earlier years, too, but this is sufficient for the demonstration)

# file preview shows a header row
deaths <- read.csv(root("demos_rstan", "trafficdeaths.csv"), header = TRUE)
head(deaths)
##   year deaths
## 1 1993    434
## 2 1994    423
## 3 1995    411
## 4 1996    355
## 5 1997    391
## 6 1998    367

First plot just the data.

deaths |>
  ggplot(aes(x=year, y=deaths)) +
  geom_point() +
  labs(y = 'Traffic deaths', x= "Year") +
  guides(linetype = "none")