• R is a programming language
• RStudio is the most widespread IDE associated to R
  
• With Python, R is one of the two top data science languages
  
• Millions of R users worldwide and the number increases rapidly
  
• Examples of use cases:
  https://github.com/ThinkR-open/companies-using-r
  https://posit.co/about/customer-stories/

• R is usually better for statistics;
  
• Python is preferred by computer scientists and people working on very Deep Learning;
  
• both are amazing for data science, graphs and simple machine learning.

Beyond statistical analyses and data science, you can do lots of things with R: reports, websites, books, applications, and slides (like these ones).
+ doc & ppt (see https://ardata-fr.github.io/officeverse/), but the latter are usually ugly.

R & RStudio can easily be combined to other languages (Python, C/C++, SQL, JavaScript, etc.).

What does “where I want you to be” mean?

Here are 2 examples of some of my students’ dynamic dashboards:

https://tmespe.shinyapps.io/Games_dashboard/
https://jitrayupunrattanapongs.shinyapps.io/dist/

My role/hope is to make you code similar objects.

But you need to start working hard early.

Knowledge is not free: you have to pay attention.

• Learning comes from YOU! More than 80% of what you will get from the course will come from your efforts. Passive listening \(\neq\) learning.
  
• Remote teaching is a major barrier! Because attention spans are short, especially in front of pedagogical videos!
• Remote teaching is not a major hurdle! Because anyway, progress will only be made by practice (on your computer, outside the video sessions).
  
• I will always be there to help. Google, stackoverflow & chatGPT are your best friends. I’m next on the list.
  
• To optimize my feedback, be as precise as possible. (best solution: send me files & code, not screenshots!)

• Large Language Models (LLMs) such as GPT & co can be useful.
  
• But the wise student should know the difference between asking for help and outsourcing.
  
• The goal is to learn to code, not to prompt!
  
• If chatGPT can do a job: you won’t get the job!
  
• Also: chatGPT makes a lot of errors. You need to (double) check.

source: https://github.com/allisonhorst/stats-illustrations

In this course, we will mostly overlook the Model part.

My only goal: that Excel becomes marginal in your workflow! 😉

1. Introduction to R/RStudio & the tidyverse
   
2. Baseline R & data structures + import/export
   
3. Plots + options
   
4. Shiny 1 - User interface & Server
   
5. Shiny 2 - UI layout (organization: tabs, rows, columns, boxes, menus, etc.)
   
6. Shiny 3 - Deployment + further options (CSS, themes & advanced tricks)
   
7. Geocomputing / Text Mining / APIs
   
8. Leveraging chatGPT / Advanced modelling / + options \(\rightarrow\) possibly to be defined together

\(\rightarrow\) Don’t hesitate to submit ideas or wishes!

One of the great strengths of R is a (very) large collection of packages.

Packages are libraries that expand the capabilities of R.

It includes:

• ggplot: the best plotting engine in the world (seriously)
  
• dplyr: for data manipulation
• tidyr: helps you work with tidy data (more on that soon)
• readr and readxl: import rectangular data files (excel, etc.)
• and more!

There are two steps: first you need to download the package (only once, like pip in Python). The packages are downloaded from servers all around the world. So we start by choosing one.

# chooseCRANmirror() to set downloading location,
chooseCRANmirror(ind = 1)    #  see getCRANmirrors() for geographical details

install.packages("tidyverse") # Install (download) package: only once

NOTE (reminder): it’s easy to install packages directly in RStudio (“Tools” tab). + .libPaths()

Then, if you want to work with it, you need to load/activate it (for every session).

library(tidyverse)            # Loading, like "import..." in Python

Sometimes, packages are not verified by the CRAN (Comprehensive R Archive Network).
They are simply hosted on Github.
To install them, use use the devtools package:

devtools::install_github("xvrdm/ggrough")

But you have to install devtools first!

• One hashtag # precedes a comment in R code.
  
• Two hashtags ## precede output from a code sequence in the slides or notebooks.

1+1 # Test!

## [1] 2

In these slides, code will appear in grey areas (rectangles).

In R, we don’t use “=” to create variables, but an arrow sign “<-”.
Though “=” works, too.

a <- 6 # This creates a variable but does not show it!
a      # If you want to see it, ask for it!

## [1] 6

b <- 11:42  # This creates a variable but does not show it!
b          # If you want to see it, ask for it!

##  [1] 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
## [26] 36 37 38 39 40 41 42

The brackets (ex: [26]) indicate the position of the first element in the row.

The diamonds database is included in the tidyverse. The head() function shows the first lines of a dataset. The tail() function shows the last lines.

head(diamonds, 4) # The number gives the amount of rows shown

One instance = one observation = one row.
One variable = one unique characteristic = one column.

• number (numerical): integer (int) or double (dbl)
• character: text (chr)
• factor: categorical, ordered (ord) or not (fct)
• boolean: TRUE or FALSE / T or F (bool)
• date: day precision (date) or second precision (time): usually starts 1970-01-01

NOTE: we are only concerned with rectangular / structured datasets.

install.packages("gapminder") # Install (download) package: only once

library(gapminder)            # Activate: each time you launch RStudio
head(gapminder, 3)            # Have a look!

The table below shows the evolution of population of countries.

pivot_wider(gapminder[c(1:4,13:16,25:28), c(1,3,5)],  # Don't look at this code!
            names_from = "country", values_from = "pop")

\(\rightarrow\) Not tidy! The columns are not VARIABLES!
(This is typically the excel format.)

The tidyverse has two functions that switch from matrix/excel format to tidy data and back:

• pivot_longer(): from matrix/excel to tidy data (wide_to_long/melt in pandas)
• pivot_wider(): from tidy data to matrix/excel (pivot in pandas)

Source: software carpentry

The reverse operation (no need for the “cols” argument this time).

pivot_wider(tidy_pop, names_from = "Country", values_from = "Population")

pivot_wider(data,
\(\quad\) names_from = variable_to_be_put_in_columns,
\(\quad\) values_from = where_to_get_values
)

Often, analyses are performed on subsets of data (query in Python).

filter(gapminder, lifeExp > 81.5) # Countries where people live long lives on average

Filters can be combined (with commas preferably, the & operator works, too).

filter(gapminder, country == "Japan", year > 2000) 

Only two observations for Japan post-2000.
NOTE: as in all languages, there are TWO EQUAL SIGNS (==) for the comparison.
One “=” is like the arrow (<-) and is used to assign values.

Sometimes, you might want to keep just a few variables to ease readability.

select(gapminder[1:5,], country, year, pop)

Use select(data, -variable) to remove variable: the minus sign!

Very often, one simple analysis will require several steps. They can be combined via the %>% or |> operators.

A fake sequence:

me %>%
\(\quad\) wake_up(time = “06:20”) %>%
\(\quad\) shower(temp = 40) %>%
\(\quad\) go_to(place = “baker”, with = “scooter”) %>%
\(\quad\) buy(item = “bread”) %>%
\(\quad\) go_to(place = “home”, with = “scooter”) %>%
\(\quad\) breakfast(drink = “hot_chocolate”, eat = “toast”, eat = “kiwi”) %>%
\(\quad\) toothbrush(duration = 2)

select(filter(diamonds, carat > 4), carat, price, clarity) # BEURK!

diamonds |>
    filter(carat > 4) |>
    select(carat, price, clarity)  # So simple!

diamonds %>% 
    filter(carat > 2, cut == "Ideal") %>%      # First we filter
    mutate(car_price_ratio = carat/price) %>%  # Then, we create a new column
    arrange(desc(car_price_ratio)) %>%         # We order the data
    select(-x, -y, -z) %>%                     # We take out some superfluous columns
    head(4)                                    # Finally, we ask for the top 4 instances

“A pivot table is a table of statistics that summarizes the data of a more extensive table.”

— Wikipedia

There are two dimensions in a pivot table:
- which variable(s) we want to analyze (gender, continent/country, size, etc.);
- which statistic we want to compute (mean, min, max, number of instances, variance etc.).

In R, these two steps are separated via two functions: group_by() and summarise()

diamonds |>
    group_by(clarity, cut) |>               # Define the variables
    summarise(avg_price = mean(price),      # Define the statistics
              max_price = max(price),
              avg_carat = mean(carat),
              max_carat = max(carat)) |>   
    head(3)

You can even pipe inside a function!

gapminder %>%
    group_by(continent, year) %>%
    summarise(avg_lifeExp = mean(lifeExp) %>% round(2)) %>%
    head(4)

The round() function rounds numbers up to some decimals.

• visualize tidyverse code: https://tidydatatutor.com
  
• equivalent for pandas: https://pandastutor.com

R is an incredibly powerful tool for data science. The preferred environment is the tidyverse. As its name indicates, the core concept is TIDY DATA!

In short, it’s just a question of functions:
- pivot_longer() and pivot_wider() to work with tidy data;
- filter(), select(), arrange() and mutate() for wrangling/manipulation;
- group_by() and summarize() for pivot tables;
- head() and tail() to see the first and last lines of a dataset.

That’s it!

• an online series of exercises with solutions!:
  https://gcoqueret.shinyapps.io/Exercises/
• the Bible of Data Science with R:
  http://r4ds.hadley.nz/
  
• Two great books on data science:
  https://ubc-dsci.github.io/introduction-to-datascience/
  https://rafalab.github.io/dsbook/
• Shiny use cases:
  https://shiny.posit.co/r/gallery/
  
• One other app (marketing):
  https://mdancho84.shinyapps.io/shiny-app/

• Programming AND learning to code is HARD:
• Ask questions so I can make it (slightly) easier.
• I can’t guess your questions, don’t be shy!
• Nothing will replace practice & making mistakes.

Tip: for your project, choose a dataset with a mixture of numerical and categorical data. It is the best combination to create a nice looking dashboard!