A gentle guide to using the {targets} package for analytic pipelines

Newcastle, UK

Nicholas Tierney

The Kids Institute, Australia

2024-09-03

Motivation

• Data analysis is interactive and messy
• How to know if it is up-to-date? Better re-run…
• Months later, how did you do things? Better re-run…
• Thinking about data analysis deserves inherent structure
• You are always collaborating with your future self
• {targets} is a way to manage complexity
  • It’s also a mindset, a philosophy.

I have 3 goals

1. Convince you {targets} is worth learning
2. Convince you that you can use it
3. For you to feel more confident writing functions

Audience

1. Who uses R here?
2. Have you used targets?

How this will all flow

• Talk / analysis is on github: njtierney/gentle-intro-targets
• Live coding (from scratch) a data analysis
• There is no “Here’s one I prepared earlier”
  • (Although I have practiced the live coding)
• Please ask questions throughout
• Ask on Canapii for deeper questions
• Also please feel free to yell out and we can transcribe/repeat
• Analysis is a little simplified
• “What questions do you have?”

How to follow along

• This was not designed as a follow along exercise
  • You are welcome to try!
• You might be best served to observe, take notes, ask questions
• Please do ask questions

Data analysis: penguins

• palmerpenguins
• Sexual dimorphism in penguins: Ecological Sexual Dimorphism and Environmental Variability within a Community of Antarctic Penguins (Genus Pygoscelis)
• Let’s do something similar, build a logistic regression, predicting sex based on penguin bill dimensions

begin_demo(timer!)

Intermission: checking in

• What things could we do to tidy up this data analysis?
• What should we do from here?
• What problems might come up in this data analysis?
• Demonstrating debugging in targets
• Reading/Writing CSV
• Saving plots

end_demo()

Takeaways: functions

• Write functions to think about solving problems
• Good functions do the following:
  • Manage complexity
  • Explain and express ideas
  • Can be individually reasoned with
  • Take iteration
• Functions can help plan out intent
• {fnmate} keyboard shortcut for writing functions helps a lot

Takeaways: {targets}

• No problem is too small for targets
• Use tflow::use_tflow() to setup a project
• Use a quarto/rmd as a scratchpad
• Keyboard shortcuts for load/make targets

Some things I might not have gotten to

• Demonstrating debugging in targets
• Reading/Writing CSV
• Saving plots

Thanks

• Will Landau
• Miles McBain
• Nick Golding
• Eric Scott

Resources

• https://books.ropensci.org/targets/
• https://milesmcbain.github.io/ssa_targets_workshop/
• https://github.com/MilesMcBain/tflow
• https://github.com/MilesMcBain/fnmate

Colophon

• Slides made using quarto
• Colours taken + modified from lorikeet theme from ochRe njtierney/njt-talks

Learning more

targets

talk link

njtierney

nicholas.tierney@gmail.com

End.