Data Science: A First Introduction

This is the source for the Data Science: A First Introduction textbook.

The book is available online at: https://datasciencebook.ca/

© 2020 Tiffany A. Timbers, Trevor Campbell, Melissa Lee

For the python version of the textbook, please visit https://python.datasciencebook.ca or the github repository at https://github.com/ubc-dsci/introduction-to-datascience-python.

License Information

This textbook is offered under the Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) License. See the license file for more information.

Development

Setup

Building the book requires Docker (instructors here: https://docs.docker.com/get-docker/)

Build locally

You must have at least 8GB of RAM (and ideally more like 16GB RAM) to build the book.

You can build the HTML version of the book on your own machine by running

./build_html.sh

in the root directory of this repository. The book can be viewed in your browser by opening the docs/index.html file.

You can build the PDF version of the book on your own machine by running

./build_pdf.sh

in the root directory of this repository. The book can be viewed in a PDF reader by opening docs/_main.pdf.

Working with RStudio (HTML only)

If you want to edit the source material and build the book using RStudio, navigate to the repository root and run

docker-compose up -d

to start up the docker container. Then open a web browser and type http://localhost:8787/. For the username enter rstudio, and for the password enter password. At any point you can render the book by running the following R code in the R console:

bookdown::render_book('index.Rmd', 'bookdown::gitbook')

When you are done working, make sure to type docker-compose down to shut down the container.

Contributing

Primary development in this repository happens on the main branch. If you want to contribute to the book, please branch off of main and make a pull request into main. You cannot commit directly to main.

The production branch contains the source material corresponding to the current publicly-viewable version of the book website.

The gh-pages branch serves the current book website at https://datasciencebook.ca.

Workflows

Book deployment

You can update the live, publicly viewable HTML book by making changes to the source/ folder in the production branch (e.g. by merging main into production). GitHub will trigger a rebuild of the public HTML site, and store the built book in the root folder of the gh-pages branch.

main deploy previews

Any commit to source/** on the main branch (from a merged PR) will trigger a rebuild of the development preview site served at https://datasciencebook.ca/dev. The built preview book will be stored in the dev/ folder on the gh-pages branch.

PR deploy previews

Any PR to source/ will trigger a build of a PR preview site at https://datasciencebook.ca/pull###, where ### is the number of the pull request. The built preview book will be stored in the pull###/ folder on the gh-pages branch.

Build environment updates

Any PR to Dockerfile will trigger a rebuild of the docker image, push it to DockerHub, and update the image tags in the build_html.sh and build_pdf.sh scripts on the PR automatically. This new build environment will be used for the PR deploy preview mentioned above.

Style Guide

General

• 80 character line limit! This is necessary to make git diffs useful
• numbers in text should be english words ("four common mistakes" not "4 common mistakes") unless there are units (40km, not forty km)
• use Oxford commas ("a, b, and c" not "a, b and c")
• "subset" should not be used as a verb
• functions in text should not have parentheses (read_csv not read_csv())
• remove all references to "course" and "student"; replace with "reader" or "you" where necessary
• make sure we have permission to use all external resources that we use
• remove all references to "clicking on things" in the HTML version of the book (e.g. "click this link to ...")
• When we introduce a new term, use **bolding** to typeset it (but only the first introduction of the term)
• for symbols as part of the text, make sure you give them their full name and surround with parentheses so that they don't "disappear" in the rest of the text. So for example, if I have a , in the text, I should do something like "here is some text about the comma (,)". Or for <-, we should do "something like this assignment operator (<-)". There are likely exceptions to this rule though.
• Book titles in the text should be typeset in italics (e.g. R for Data Science)

Code blocks

• Use the knitr label format ##-[name with only alphanumeric + hyphens] where the ## is the 2-digit chapter number, e.g. 03-test-name for a label test-name in chapter 3
• Make sure to get syntax highlighting by specifying the language in each code block:

  ```r
     code
  ```
  

  not

  ```
    code
  ```
  (similar for `html` where needed)
  

• always use |> pipe, not %>%
• anywhere we specify a grid of tuning values, don't just do grid = 10; actually specify the values using seq or c(...)
• do not end code blocks with head(dataframe); just use dataframe to print
• set.seed once at the beginning of each chapter
• use "double quotes" for strings, not 'single quotes'
• make sure all lines of code are at most 80 characters (for LaTeX PDF output typesetting)
• pass code blocks through styler (although must obey the 80ch limit)
• use slice, slice_min, slice_max (not top_n)
• just pull(colname), don't select first

Section headings

• All (sub)section headings should be sentence case ("Loading a tabular data set", not "Loading a Tabular Data Set")
• Make sure that subsections occur in 1-step hierarchies (no subsubsection directly below subsection, for example)
• Make sure that {-} is used wherever unnumbered headings are required

Choose an appropriate table of contents depth via (example has depth 2 below, which is a good default)

bookdown::gitbook:
    toc_depth: 2

Learning objectives

• when saying that students will do things in code, always say "in R"
• "you will be able to" (not "students will be able to", "the reader will be able to")

Captions

• captions should be sentence formatted and end with a period
• If you have special characters (particularly underscores, quotation marks, plus signs, other LaTeX math symbols) make sure to separate the caption out of the code chunk like so

  (ref:blah)
  
  \`\`\`
  {r blah, other_options}
  code here
  \`\`\`
  

Equations

• make sure all equations get capitalized labels ("Equation \@ref(blah)", not "equation below" or "equation above")

Figures

• make sure all figures get (capitalized) labels ("Figure \@ref(blah)", not "figure below" or "figure above")
• make sure all figures get captions
• specify image widths of pngs and jpegs in terms of linewidth percent (e.g. out.width="70%"), for plots we create in R use fig.width and fig.height.
• center align all images via fig.align = "center"
• make sure we have permission for every figure/logo that we use
• Make sure all figures follow the visualization principles in Chapter 4
• Make sure axes are set appropriately to not inflate/deflate differences artificially where it does not compromise clarity (e.g. in the classification chapter there are a few examples where zoomed-in accuracy axes are better than using the full range 0 to 1)
• Fig size for bar charts should be: fig.width=5, fig.height=3 (an exception are figs 1.7 & 1.8 so that we can read the axis labels)
• cropping width for syntax diagrams is 1625 (done using image_crop)

Tables

• make sure all tables get capitalized labels ("Table \@ref(blah)", not "table below" or "table above")
• make sure all tables get captions
• make sure the row + column spacing is reasonable
• Do not put links in table captions, it breaks pdf rendering
• Do not put underscores in table captions, it breaks pdf rendering

Note boxes

• note boxes should be typeset as quote boxes using > and start with Note:

Bibliography

• do not put "et al" or "and others"; always use the full list of authors, BibTeX will choose how to abbreviate
• read https://trevorcampbell.me/html/bibtex.html and make sure our bib follows this convention

Naming conventions

• K-means (not $K$-*, K means, Kmeans)
• K-nearest neighbors (not $K$-*, K nearest neighbors, K nearest neighbor, use US spelling neighbor not neighbour). Note that "K-nearest neighbor" is not the singular form; "K-nearest neighbors" is
• K-NN (not $K$-*, KNN, K NN, $K$NN, K-nn)
• local repository (not local computer)
• package (not library, meta package, meta-package)
• data science (not Data Science)
• data frame (not dataframe)
• data set (not dataset)
• scatter plot (not scatterplot)
• bar plot (not bar chart)
• capitalize all initialisms and acronyms (URL not url, API not api, K-NN not k-nn)
• response variable (not target, output, label)
• predictor variable (not explanatory, feature)
• numerical variable (not quantitative variable)
• categorical variable (not class variable)

Punctuation

• emdashes should have no surrounding spaces. This kind of typesetting—which is awesome—is correct! and Typesetting with spaces around em-dashes — which is bad — is not correct
• make sure \index commands don't break punctuation spacing. E.g. This is an item \index{item}; it is good will typeset with an erroneous space after item, i.e. This is an item ; it is good

Common typos to check for

• RMPSE: should be RMSPE
• boostrap: should be bootstrap

Use American spelling

Generally the book uses American spelling. Some common British vs American and Canadian vs American gotchas:

• o vs ou: neighbor and color (not neighbour and colour)
• single vs double ell: labeling and labeled (not labelling and labelled)
• z vs s: summarize (not summarise)
• c vs s: defense (not defence)
• er vs re: center (not centre)

Whitespace

We need a line of whitespace before and after code fences (code surrounded by three backticks above and below). This is for readability, and it is essential for figure captions.

PDF Output

These are absolute last steps when rendering the PDF output:

• Look for and fix bad line breaks (e.g. with only one word on the next line, orphans, and widows)
• Look for and fix bad line wraps in code and text
• Look for and fix bad figure placement (falling off page, going over the side)
• Look for and fix large whitespace sections where LaTeX doesn't want to break the next paragraph (usually \allowdisplaybreaks helps)
• Fix incorrect indenting. LaTeX will indent for a new paragraph if there is an extra whitespace line, so these should be deleted if no paragraph break is desired.
• Look for ?? in the PDF (broken refs)
• Look in the index for near-duplicates, and merge if needed
• Look for / fix raw LaTeX code (search for backslash and curly brace in the final PDF)
• Make sure the 3D figures (and the text around them that refers to clicking and dragging) are properly modified for the PDF output
• Make sure all markdown label-replaced URLs (of the form [blah](url)) will make sense in the hardcopy book version (i.e. nothing like "click this"). Many links appear in the additional resources: make sure the text-replacement of the URL contains enough information for someone to find the resource (without being able to click the link)

HTML Output

• Look for broken references (I think these end up as ??)
• Look for uncentered images