Data Literacy with R for Students of Humanities

Welcome

Target group:
- scholars, no programming background
- supposed to be hard!!!
No question is stupid. Please repeat is excellent!
- You are asking for the shy ones.
- Teacher gets easily wrong or incoherent, like most people.
Hazard loud guesses.
- Often legit assumptions stumble on system-inherent idiosyncrasies.
- You are hardly alone thinking so.

Welcome

RStudio on cloud

https://aic.ufal.mff.cuni.cz/jlab/

Learning materials on github

https://github.com/ufal/R_BEGINNERS_SHORT

This course in the course catalogue

https://is.cuni.cz/studium/eng/predmety/index.php?do=predmet&kod=NPFL112

DataCamp

https://www.datacamp.com/users/sign_in

R vs. Python for Data Science

Which is preferred in your research domain?
- NLP, DH: both
R probably best at tabular data (manipulation, plotting)
Nice comparison at statology.org:

https://www.statology.org/r-vs-python-for-data-science-a-comprehensive-comparison/

Mind map of R

Core concepts, links to DataCamp classes https://orgpad.info/s/SKpWCNY4P7w

The Big Picture

Base R vs. “dialects”

Base R

Data structures

The signature data structure of R is the vector: ordered sequence of observations (i.e. position matters).

Each observation = vector element. A single observation = a vector with one element. Factor = a vector of character strings that has been told that the strings are values of a categorical statistical variable, so it immediately records for R the frequency of each value, without calling a counting function every time.

When vectors and factors are equally long, they can be adopted as columns into a data frame. These are the structures you are going to need regularly. Other important structure is list. It can store any other structures as its elements. For instance, you can have a list of several data frames, vectors, and other lists with anything inside.

Matrix is not much relevant for this course. It is actually a vector but it has an additional dimension of columns, which makes it similar to a data frame. However, all matrix columns must be the same data type, whereas in data frame, each column in a data frame can be a different data type (strings, numbers, Boolean values).

Functions and programming structures

A function is like a machine with a given purpose. Imagine a microwave oven. All it does is heating up, but the results of heating up depend on what you input and what time and intensity you select: Cold water -> boiling water, corn -> popcorn, … You get rubbish when you put in something unintended or select wrong intensity or time. Unlike a microwave, you cannot destroy a function though.

Control structure Loop. You can call a function (or a larger piece of code) on a series of input objects, and the code will process one after another.

Control structure Condition: It says: “If this condition is met, run this piece of code”. For instance, your piece of code takes a number and divides it by two. When you feed it a string instead of a number, it throws and error and stops your script. With a condition saying “do this only to numbers”, you make sure that an accidental string would not crash your script.

While loop and conditions are programming universals, base R has a specific family of functions to avoid loops. They are have apply in their name. You will see them a lot in base R code. They have a sligthly different syntax than regular functions, so you have to learn about them separately. When you are going to work mainly on data frames (tables), you would probably benefit from learning to use the corresponding functions from the purrr library, which is a “dialect” of R.

There is one more notable cell in the image: Iconic data sets. All libraries come with some documentation of their functions, and most demonstrate them on me on data sets that R makes available to all users - built-in data sets. Each new library can add its own built-in data sets. Most R users are familiar with quite a few such built-in data sets. These data sets come from diverse domains from botany to flight schedules.

Tidyverse

https://www.tidyverse.org/

In our course, we are going to focus on one modern R “dialect” for data manipulation, the tidyverse. It is a family of software libraries that has a fabulous library for plotting, too. Here you see icons of the tidyverse libraries. The tidyverse is associated with the name Hadley Wickham, the initial author, but it became so popular that it has grown to an entire ecosystem of complementary libraries with many authors and contributors. The core idea of tidyverse is tidy data, which means each table row represents one observation and each column represents one variable, and once you stick to this data philosophy, you can use an incredibly user-friendly and efficient toolkit.

So when you work with data frames, you mostly need to filter away some rows or select just some columns, or to add another column based on a computation on other columns or something else. Or you have the data scattered in several different tables and want to merge them into one, just as if you query a relational database. And this is what the dplyr library is for. tidyr is a library that helps you make your data comply with the tidy principles. stringr makes it easy for you to operate on strings (text search, replace, switch case etc.). readr is for reading files into R objects and saving R objects to files, e.g. a data frame to a .csv file. lubridate is for dates and times. Without it, they are just strings. When you need e.g. to compute duration from two dates and times, call lubridate. ggplot2 is the plotting library.

The last three libraries are not so obvious at this moment, but for the sake of completeness: purrr processes things in lists. forcats works with factors (those vectors that know that they are categorical variables and count frequencies of their values). tibble adjusts base R data frames to fit better with tidyverse packages.

R for Data Science

Reporting

I will first of all mention Reporting. It is admittedly not the most important programming skill, but it can help you keep great notes alongside your code right from the start of the course, so you can re-run the scripts and edit the comments any time.

There are interactive reporting tools in RStudio, which are based on a library called knitr (outside tidyverse). You write your text like plain text with a set of simple formatting tags. This format is called Markdown and you may know it from note-taking apps such as Notion or Obsidian. R uses a slightly modified “flavor” of Markdown called RMarkdown, in which you can combine executable code and free text. You can even insert images and other external objects.

RStudio will offer you three reporting tools: RMarkdown, RNotebook, and Quarto. All of them generate nice reports and slides using markdown and knitr. All are good and give a similar user experience, but Quarto is newest and likely to replace the older ones in RStudio, so it is probably wise to stick to Quarto.

Just on the margin: reporting is more than reports you can print or run on your computer. Other popular formats are interactive websites and dashboards, but these are beyond our current scope.

Note keeping tip

How to keep notes in RStudio

Source file of this presentation: 01_Introduction.qmd.
Edit your local copy to add your notes in Markdown.
To preserve the slides structure, write your notes inside ::: notes :::
You can render it as pdf/html even if you ruin the slides rendering.
How to keep the presentation structure
- https://quarto.org/docs/presentations/

To keep notes in RStudio, you open a .qmd or .Rmd file. I write my teaching materials in Quarto and my default output is revealjs slides. I write the accompanying text into a field called speaker’s notes. It is not a Markdown but a revealjs functionality. They are hidden to the audience during a slide show, but you see them when I render the same content in html or pdf, including the tags. A block of notes like this one must always have one empty line (with its own number) before and one after, otherwise the notes appear in your slide along with the tag.

You can also include latex commands and html code into markdown documents for better styling, but we will not learn this in this course. The goal here is that you have reproducible notes, and that you will get with the markdown template you see when you open a new markdown file. Here is the user guide to Quarto: https://quarto.org/

Data Literacy with R for Students of Humanities

Welcome

Welcome

R vs. Python for Data Science

Mind map of R

The Big Picture

Base R vs. “dialects”

Base R

Data structures

Functions and programming structures

Tidyverse

R for Data Science

Reporting

Note keeping tip

Data manipulation - the core skill

Data frame wrangling

Data import: `readr`

Processing text (strings)

Extracting external data

Programming

That’s R!

Data Literacy with R for Students of Humanities

Welcome

Welcome

R vs. Python for Data Science

Mind map of R

The Big Picture

Base R vs. “dialects”

Base R

Data structures

Functions and programming structures

Tidyverse

R for Data Science

Reporting

Note keeping tip

Data manipulation - the core skill

Data frame wrangling

Data import: readr

Processing text (strings)

Extracting external data

Programming

That’s R!

Data import: `readr`