• How I came to be teaching R
• What I've learnt on the way
• Hints and tips when teaching / learning R

• Previous BSC in Computer Science
• Data Analysis Msc at York
• Statistics module taught using R
• PDF lectures with practicals afterwards

• We've come a long way!

• ~2012/2013 (?)
• A pre-requisite to the courses I was already teaching
• Established and developed over a number of years
• Materials still online

• Plotting at the end of the course
  • most important part for some people
  • often skimmed over due to time constraints
• Datasets too hard
• Stats too hard
• Functions too hard

• Plotting at end of the course
  • moved much earlier
  • but something has to come at the end
• Datasets too hard
• Stats too hard
• Functions too hard
  • removed from syllabus
  • were we probably trying to fit too much in

• Tempting to use datasets that participants will use day-to-day
  • but can sometimes be complicated to explain
  • focus too much on the data and not the R code
  • not everyone will have the same background
• Generic dataset sometimes better
  • ozone / weather dataset
  • "patient" weights / heights
  • (gene expression dataset)
• Huge list of datasets with downloads thanks to Vincent Arel Bundock

• People sometimes expect an R course to be about stats
• No-win situation
  • cut-back the stats and people want more
  • add more stats content and people complain it's too much :/
• (Personal opinon) Learning R and learning stats are separate activities
  • can either focus on R and neglect interpretation of stats, or vice-versa

• Cover the use of t.test
  • doing paired t-test vs independent t-test not much harder (or wilcox.test)
  • difficulty is knowing what test to apply
• For basic stats, we developed some Shiny apps
  • focus is on interpreting data and results
• Signpost to other stats courses or resources that assume R

http://cambiotraining.github.io/r-intro/

• Introduced reproducibility and R markdown
• Slides generated with R markdown vs powerpoint
  • easier to update
• Eventually switch to HTML notes
  • no more handouts
• Good reviews, but not quite satisfied yet

• What is best to teach novices?
• My preconception (at the time) that everyone learns base R first and then learns tidyverse afterwards
  • probably based on how I learnt R!
• Not all packages use ggplot2 for plotting
  • some types of data (e.g. Genomics) not traditionally "long" format
• (Spoiler alert - I eventually decided to teach tidyverse rather than base)

• Online here
• Assumes some basic R knowledge
• Covers dplyr and ggplot2 in detail
• Simple simulated dataset of patient demographics

• New intermediate R is getting rave reviews
• Using best-practice tools
  • dplyr, ggplot2
  • R markdown, github
• So everyone is using R now?
  • why not
  • (those of a nervous disposition might want to look away)

• Data formatting
• Have to accept that people are going to use Excel
• Teach them to use it safely
  • no colours
  • one piece of data per column
  • no spaces
  • consistent groups (Male, Female, M, F, male…)
  • etc
• What about version control?

https://carpentries.org/

• Live coding
• Sticky notes
  • Yellow = Ready to move on
  • Pink = Help me
  • Also used to get feedback
• Instructor training programme

http://sbc.shef.ac.uk/training/r-introduction-2019-03-29/

• no more base R!
• live coding all the way
  • except 10 minute introduction
• using gapminder data

suppressPackageStartupMessages(library(tidyverse))
read.csv("gapminder.csv") %>% 
  filter(continent!="Oceania") %>% 
  ggplot(aes(x=as.factor(year),y=lifeExp,fill=continent)) + geom_boxplot() + 
  facet_wrap(~continent)

• Plenty of materials for re-use
  • carpentry materials can be used (with caveat)
• We are building a network of trainers / demonstrators
  • interested in carpentry instructor training? let us know
• Don't try and do too much
  • focus on Core skills, and do them well
• Do something exciting early on
  • a cool plot, filter some data
• Choose dataset carefully
  • unless you can guarentee background of participants, choose something generic. e.g. gapminder
  • maybe add an extra section analysing a domain-specific dataset

• Setup on participant machines can be a headache
  • Tempting to use a computer room with everything pre-installed
  • But will they be able to use it afterwards?
• Send installation instructions in advance
  • offer drop-in sessions to help with install?
• Could use Docker / Virtual Machine
  • But could be as complicated to setup as R itself
  • Cloud compute is an option (AWS etc)

• Practice, practice, practice
  • if you don't have your own data, find some fun datasets to play with
• Read the blogs / twitter
  • R-bloggers
• Persevere

"The bad news is that when ever you learn a new skill you're going to suck. It's going to be frustrating. The good news is that is typical and happens to everyone and it is only temporary. You can't go from knowing nothing to becoming an expert without going through a period of great frustration and great suckiness." - Hadley Wickham, useR 2014