Today will be a fast tour of modern R.

• R is a moving target

• Focus on “Tidyverse”
  • dplyr, ggplot2, tidyr, etc
  • Mostly written by Hadley Wickham

S - 1976

R - 1993

Exploratory

Complex experimental designs, large \(n\) and/or large \(p\).

Understanding is a cyclic process of exploratation.

Integrative

In biology you may want to join together many different views of a process: DNA, RNA, epigenetics, proteins, metabolome, cell morphology, …

Your data can be viewed in the context of many other data sets.

In biology, reference genome and gene annotations are the key to joining different types of data.

See also: The “R for Data Science” book

Load, tidy, normalize and transform data

• Tidy data makes visualization and modelling fast and easy.
• Correctly normalized and transformed data brings out the signal in visualizations, and necessary for correct modelling.

Key packages: readr, tidyr, dplyr

Visualization, exploration

The greatest value of a picture is when it forces us
to notice what we never expected to see.

– John Tukey

• Suggests the need for normalization or transformation.
• Shows if any of the data is poor quality.
• Shows unexpected things in the data.
• Informs what relationships are important in the data for any modelling and statistical testing.

Key packages: ggplot2, shiny

Summarization, modelling, statistical testing

• Shrinks large data sets for more manageable visualization.
• Confirm what can be seen in visualization, tells us what isn’t real.
• What hypotheses does the data support and reject?
• Failures should prompt a rethink of normalization and tranformation, and further visualization to understand the data.

Key packages: dplyr

Base R functions: mean, min, max, sd, lm, glm, anova, …

Specialized packages: too many to name

Note: Model fitting and hypothesis testing won’t be covered today.

Tidy data doesn’t mean tidy for a person to read, it means the easiest form for the computer to work with.

• only use data frames
• each row is a single unit of observation
• each column is a single piece of information
• each column is a distinct kind of information

Similar to database design.

The experimental design is in the body of the table alongside the data, not in row names or column names.

Example from: Wickham, H. (2015) Tidy data. The Journal of Statistical Software, vol. 59.

If every step of your analysis is recorded in an R script, with no manual steps:

• you have a complete record of what you have done
• easy to run entire script with test data
• changes easily tested, poor early decisions easily fixed
• today’s big project becomes tomorrow’s building block

Open, reproducable data science.

Show your code. It may surprise you who picks it up and runs with it.

• Document analysis in Rmarkdown

• Ultimately, share useful functions in R packages
  • Version control (eg git, GitHub or GitLab)
  • Tutorial and reference documentation
  • Testing
  • devtools::check

• For reproducibility, record which versions of packages you used
  • sessionInfo()

Rmarkdown documents

Programming

Tidying data and visualizing it

Sharing data interactively

Also in notes

Working with DNA sequences and genomic feature data