Live coding: your first Quarto report

Session 1 β€” Applied Data Science 2026

Author

Claudius GrΓ€bner-Radkowitsch

Published

14 09 2026

This is the document we build together in the session, written from an empty file step by step. Use it afterwards as a model for your own reports.

1 Step 0: set up a project

Before writing a single line of code, give your work a home.

ImportantDo this now β€” this should be the first step for each new project of yours
  1. In RStudio: File β†’ New Project β†’ New Directory β†’ New Project
  2. Name it applied-data-science
  3. Put it somewhere you will find it again β€” not in Downloads or Desktop
  4. Click Create Project
  5. Using New Folder in the Files pane, create data, R and Output
  6. Download hotels-vienna.csv from the session page and save it into data/

You should end up with this:

applied-data-science/
β”œβ”€β”€ applied-data-science.Rproj   ← created for you; never edit it
β”œβ”€β”€ data/                        ← raw data, read-only. Never edit by hand.
β”‚   └── hotels-vienna.csv
β”œβ”€β”€ R/                           ← .R scripts (we start using this in session 2)
β”œβ”€β”€ Output/                      ← rendered reports, exported figures and tables
└── session01.qmd                ← the document we are about to create

The rule for each folder: data/ goes in and never changes; R/ holds reusable code; Output/ holds anything produced by your code, and should be safe to delete because you can always regenerate it.

NoteHow you organise the semester is up to you

I keep one project for the whole course and you are welcome to copy that. One project per session works just as well, and nothing in this course depends on which you pick. Reports are the exception β€” each report gets its own project, for reasons we come to in session 3.

Two things matter whichever way you go:

  • The .qmd you are working on sits at the top of the project, with data/ beside it. Every session reads its data as data/something.csv, so if you bury the document one folder deeper, the path stops working. Session 3 shows the proper fix for that; until then, keep it flat.
  • If you keep one project, name the files so that future-you can tell them apart: session01.qmd, session02.qmd, and so on. By December there will be a dozen of them, and three files called Untitled1.qmd will tell you nothing.

RStudio restarts and you are now inside the project. Two things changed: the top-right corner shows the project name, and R’s working directory is the project folder.

That second one matters more than it looks. It means you can write data/hotels-vienna.csv instead of C:/Users/yourname/Desktop/stuff/final_v3/data/hotels-vienna.csv β€” and your code will still work on someone else’s computer.

TipThe rule

Never write a file path that starts with C:/ or /Users/. If you do, your analysis only runs on your machine, which means it is not reproducible, which means it is not evidence.

2 Step 1: the console vs. the script

Type this in the Console (bottom left) and press Enter:

2 + 2
[1] 4

R answers. Now something more useful β€” store a value:

revenue <- 12500
costs <- 9800

Nothing is printed. Look at the Environment pane (top right): the two objects are now there. Ask for them:

revenue - costs
[1] 2700

The arrow <- is how you name things. Read it as β€œgets”: revenue gets 12500.

WarningConsole work disappears

Everything you type in the console is gone when you close RStudio. If you want to keep it, write in a file instead!

3 Step 2: create the Quarto document

File β†’ New File β†’ Quarto Document. Give it a title, choose HTML, click Create. Save it as session01.qmd.

RStudio gives you a file with three kinds of content:

Part What it is How to spot it
YAML header Settings for the whole document Between the two --- lines at the top
Text Your prose, in Markdown Plain writing
Code chunks R code that gets run Between ```{r} and ```

3.1 The YAML header

Ours looks like this:

---
title: "What does a hotel room cost in Vienna?"
author: "Your Name"
date: today
format: html
---

That is the minimum. format: pdf would produce a PDF instead.

3.2 Writing text

Markdown is plain text with a few marks:

You write You get
# Heading A big heading
## Subheading A smaller one
**bold** bold
*italic* italic
- item a bullet point

That is genuinely most of it.

4 Step 3: your first code chunk

Insert a chunk with Ctrl/Cmd + Alt + I, or type it out:

library(tidyverse)
library(flextable)

This loads the tidyverse, a collection of packages we use all semester. It also loads flextable, which makes tables look nicer in a report.

NotePackage vs. library

A package is installed once, with install.packages("tidyverse"). It is loaded every session, with library(tidyverse). Installing is buying the book; loading is taking it off the shelf.

5 Step 4: load the data

hotels <- read_csv("data/hotels-vienna.csv")

We will take read_csv() apart properly next week. For today: it reads a comma-separated file and gives you a table.

NoteWhere this data comes from

428 hotels in Vienna, price for one night in November 2017. From BΓ©kΓ©s & KΓ©zdi (2021), Data Analysis for Business, Economics and Policy, licensed CC BY-NC-SA 4.0. Each row is one hotel.

6 Step 5: look before you leap

Three questions to ask of any new dataset.

How big is it?

nrow(hotels)
[1] 428
ncol(hotels)
[1] 24

What is in it?

names(hotels)
 [1] "country"            "city_actual"        "rating_count"      
 [4] "center1label"       "center2label"       "neighbourhood"     
 [7] "price"              "city"               "stars"             
[10] "ratingta"           "ratingta_count"     "scarce_room"       
[13] "hotel_id"           "offer"              "offer_cat"         
[16] "year"               "month"              "weekend"           
[19] "holiday"            "distance"           "distance_alter"    
[22] "accommodation_type" "nnights"            "rating"

What do the numbers look like?

summary(hotels$price)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   27.0    83.0   109.5   131.4   146.0  1012.0

The $ picks one column out of the table, so hotels$price is the price column on its own.

TipRead that summary properly

Median €109.50, mean €131.40. The mean sits well above the median, so the distribution is right-skewed β€” a few very expensive hotels pull the average up. For prices, incomes and firm sizes this is the norm, and the median is the more honest summary.

Also note the maximum: €1012. Hold that thought.

7 Step 6: your first chart

ggplot(
  data = hotels,
  mapping = aes(x = distance, y = price)
) +
  geom_point(alpha = 0.5) +
  theme_classic()
Figure 1: Price against distance from the city centre, 428 Vienna hotels.

Three things are going on, and they are the same three things in every ggplot:

  1. data = β€” which table
  2. mapping = aes(...) β€” which column goes on which axis
  3. geom_*() β€” how to draw it

alpha = 0.5 makes points semi-transparent so we can see where they overlap.

theme_classic() is one line making the plot look more decent.

7.1 Interrogating the chart

The headline is unsurprising: closer to the centre, more expensive. The interesting part is everything else.

Who is that hotel at €1012?

hotels |>
  filter(price > 500) |>
  select(price, stars, distance, accommodation_type, neighbourhood) |> 
  flextable() |> 
  theme_vanilla()

price

stars

distance

accommodation_type

neighbourhood

513

5.0

0.3

Hotel

Innere Stadt

806

5.0

0.3

Hotel

Innere Stadt

544

5.0

0.5

Hotel

Innere Stadt

511

3.5

0.5

Apartment

Innere Stadt

527

5.0

0.4

Hotel

Kaerntner Strasse

587

5.0

0.7

Hotel

Leopoldstadt

1,012

3.0

1.0

Hotel

Neubau

A three-star hotel at €1012, when the most expensive five-stars are around €800. That is almost certainly a data-entry error, and we found it by looking at a picture rather than by running a test.

And who is 13 km out?

hotels |>
  filter(distance > 6) |>
  select(city_actual, price, distance, stars) |> 
  flextable() |> 
  theme_vanilla()

city_actual

price

distance

stars

Fischamend

59

13.0

3

Schwechat

92

11.0

3

Schwechat

166

11.0

4

Schwechat

137

7.7

4

Schwechat

99

7.7

3

Schwechat

140

7.8

4

Schwechat

80

6.3

3

Schwechat

91

6.7

3

Vienna

134

6.9

3

Vienna

130

6.6

3

Voesendorf

89

7.0

4

Voesendorf

92

7.1

4

Twelve properties sit more than 6 km out β€” and for ten of them the city_actual column does not say Vienna. Schwechat is the airport town, Fischamend and Voesendorf are neighbours. β€œDistance to the city centre” means something quite different for an airport hotel, so those ten should come out of most analyses.

The other two genuinely are in Vienna, just far from the middle. They stay.

TipFilter on the reason, not on the symptom

It would be easy to write filter(distance < 6) and move on. That would also throw away two perfectly good Vienna hotels, and it would silently break the moment someone adds a hotel 6.5 km out in the 22nd district.

The thing we actually object to is that ten properties are not in the city, so that is what we should filter on:

hotels |>
  filter(city_actual != "Vienna") |>
  nrow()
[1] 10
ImportantThis is the actual job

Nobody told us the data had these problems. We found them in the first five minutes by plotting it and asking who the odd points were. Do this every time, before you model anything.

8 Step 7: adding a second variable

Colour by star rating:

ggplot(
  data = hotels,
  mapping = aes(x = distance, y = price, colour = factor(stars))
) +
  geom_point(alpha = 0.7) +
  labs(
    x = "Distance to city centre (km)",
    y = "Price per night (EUR)",
    colour = "Stars",
    title = "Central hotels cost more β€” but they are also fancier"
  ) +
  theme_classic()
Figure 2: The same data, coloured by star rating.

Look at the colours near the left edge. The expensive central hotels are also the high-star ones.

NoteThe question for next week

Is the centre expensive because it is central, or because that is where the good hotels happen to be?

The chart cannot tell us. Session 2 gets us most of the way, and session 8 finishes the job.

9 Step 8: render

Click Render (or Ctrl/Cmd + Shift + K).

Quarto runs every chunk from top to bottom in a fresh R session, and writes an HTML file next to your .qmd.

WarningWhy rendering fails

Rendering starts from nothing. If your code only works because you typed something into the console an hour ago, it will fail here.

This is a feature. It is the machine checking that your analysis actually reproduces. It is also exactly what the automated check on your reports will do.

10 Step 9: useful chunk options

Options go at the top of a chunk, after #|:

Option Effect
#| echo: false Run the code, hide it β€” show only the output
#| eval: false Show the code, do not run it
#| warning: false Suppress warning messages
#| fig-cap: "..." Add a caption to a figure
#| label: fig-name Name the figure so you can refer to it

Example β€” a chart with no code shown:

Where the prices sit.

Almost every hotel sits on the left. A handful stretch the axis out to €1012, and they squash everybody else into a narrow band β€” the same reason the scatter earlier looked like a dense clump with a few strays.

Prices behave like this almost everywhere: a hotel can be ten times the average, but never a tenth of it. The usual fix is to plot price on a logarithmic scale, which spreads the crowded part out and pulls the tail in.

We are not doing it today, deliberately. On a log scale the €1012 hotel stops looking like a mistake and starts looking like an ordinary expensive hotel β€” and finding that mistake was the whole point of this session. Session 5 picks this up properly by discussing the cases where you must transform your data using logarithms.

11 Recap

You now have all of it:

  • A project so paths work everywhere
  • A .qmd file mixing text and code
  • Chunks that run R
  • A rendered HTML report you could send to someone
  • And two data problems nobody warned you about
TipIf you remember three things
  1. Write in the Source pane by default, not the console
  2. Never use absolute file paths
  3. Plot the data before you believe anything about it