1 Introduction

Important

The post is not finished. Stay tuned for updates!

Hi there and welcome to my first project for the blog! The topic is a sad one, but I would like to explain why I decided to start with this. I was trying Shiny for different dashboards, but I wasn’t satisfied to learn using the classic examples. Unfortunately, italian crime news suffocated the public debate with a case of homicide, in which the victim is a young women. The public debate was focusing so baaaaaadly on the concept of femicide and the data, that I decided to clear the situation with a simple dashboard.

First of all, at this link you can find the dashboard published using shinyapps.io. Also at this link you can find the Github repo. As you can see, even if it is the main branch, there are some details which are not uber perfect, but that don’t interfere with the code.

Now lets jump into the detail of how to create a Shiny dashboard!

2 The structure

You can create a very basic shiny dashboard with RStudio by creating a new project and select Shiny Application.

By default, a Shiny Application consists of a server.R file, in which there a 2 functions:

ui(), which defines the User Interface.
server(), which defines the back-end of your application.

But wait!? My code is different! This is because when a shiny app become bigger, it is useful to separate the code in different files:

global.R file to declare all the variables and all the packages.
ui.R file for the front-end.
server.R for the back-end.

Doing so has its pros and cons: each part of the application has its own raison de vivre and is self contained; still, you need to be very precise in commenting each part of the code, so you can remember all the interconnection of the ui with the server file.

could be nice to have a diagram of the 3 files which convergence in the app.

3 The code

In this part we are going to analyze how the app is composed. I would like to follow the files flow.

3.1 global.R

As said, this file is our usual R Script file. First thing first, we import our libraries:

Code

# wrangling
library("tidyverse")
library("readr")
library("stringr")
library("dplyr")
library("magrittr")
library("forcats")
library("lubridate")
library("writexl")
library("eurostat")
# plotting and dashboarding
library("shiny")
library("shinythemes")
library("ggplot2")
library("plotly")
library("scales")
library("RColorBrewer")
library("waiter")
# connecting and other
library("rsconnect")
library("markdown")

Lots of packages! The division is merely to remember how everything is managed and because I have OCD for this type of things.

I want to focus on some packages:

tidyverse, we all know it. As you can see, I also imported lots of single packages which compose the tidyverse, because I was getting errors of missing methods.
eurostat, which lets data flows from the eurostat website to my dashboard. This also lets the dashboard automatically updates when new data is available.
scales, to nicely scaling my x and y axis.
RColorBrewer, because I wanted to have a colorblind safe dashboard, even tough I am not.
waiter, for nice waiting images while the dashboard is loading.

Now we can focus on the data importing and wrangling. For this, the eurostat library does the job. Lets focus on the crim_hom_vrel dataset.

Code

# search in eurostat db
homicide <- search_eurostat("homicide")

# import data to variable
crim_hom_vrel <- get_eurostat("crim_hom_vrel", time_format = "date")

# convert all observations to understandable data
crim_hom_vrel <- label_eurostat(crim_hom_vrel)

# label_eurostat_vars(crim_hom_vrel)

# order data by country and date for time series purpose
crim_hom_vrel <- crim_hom_vrel %>% 
  arrange(geo, time)

crim_hom_vrel_grouped <- crim_hom_vrel %>% 
  dplyr::group_by(geo, time, sex, pers_cat, unit) %>% 
  dplyr::summarise(values_grouped = sum(values), .groups = "drop") %>% 
  filter(unit == "Number") %>% 
  arrange(geo, time, sex)

Everything pretty simple. I would like to highlight something about the dplyr::group_by and dplyr::summarise. As you can see, after having grouped and summarized, I need to drop the groups with the method .groups = "drop". With dplyr v.1.1.0 we can do the same with the help of the .by method in summarise.

Code

crim_hom_vrel_grouped <- crim_hom_vrel %>% 
  dplyr::summarise(
    values_grouped = sum(values),
    .by = c(geo, time, sex, pers_cat, unit)
    ) %>% 
  filter(unit == "Number") %>% 
  arrange(geo, time, sex)

Copying from the dplyr website the differences between .by and group_by() are:

.by	group_by()
Grouping only affects a single verb	Grouping is persistent across multiple verbs
Selects variables with tidy-select	Computes expressions with data-masking
Summaries use existing order of group keys	Summaries sort group keys in ascending order

Last part is all about colors.

Code

# brewer.pal(11, "RdYlBu")
palette <- c("#A50026", "#D73027", "#F46D43", "#FDAE61", "#FEE090", "#FFFFBF", "#E0F3F8", "#ABD9E9", "#74ADD1", "#4575B4", "#313695")

palette_crim_hom_vrel_grouped <- rep(
  palette,
  length.out = crim_hom_vrel_grouped$geo %>% str_unique() %>% length()
  )

Here I defined the palette using ColorBrewer. Using rep I replicated the 11 colours for the length of the unique geo values.

Eurostat Homicide Data

1 Introduction

2 The structure

3 The code

3.1 global.R

3.2 ui.R