28th January 2022
https://is.gd/reproducibility

Contents

Not about the birds and the bees! Instead reproducibility


Layout:

1. Some stories

2. A task for you

3. Another story

4. Some advice

Story number 1: Climate change and pollen

Story number 2: Radiocarbon dating

Radiocarbon calibration

Part 2: Reproducibility and the IPCC; a task

Confidence in the IPCC

Your mission…

  1. Download this file from GitHub: https://is.gd/IPCC2022
  2. There are 62 pages, so run sample(1:62, 1). Then count the number of references and do sample(1:n_ref, 1) to pick a random paper and copy the DOI
  3. Search the paper for http. See if there is code or data linked to the article that can be downloaded
  4. Copy across the URLs of the data/code links and…
  5. Fill all this in the Google form: https://is.gd/IPCCform

Repeat as many times as you want

Another story: my attempt at reproducibility done properly

Back in time

A stable isotope mixing model

First publications

A social media ‘strategy’

Environmetrics and doing it wrong

How it’s used

  • Management and potential collapse of important microscopic plankton in the Ionian Sea,
  • Whether American Black Bears are consuming human crops,
  • The impacts of predator/prey removal and the new EU Common Fishery policy,
  • The effects of land use on living beings in river streams and the link between storm-water run-off and pollution,
  • How the diets of individual Neanderthals differed from others,
  • The metabolic rate of Bottlenose dolphins,
  • The life cycle of Uruguayan Green turtles,
  • The increase of trace heavy metals in seagrass leaves

GitHub pages and manuals

Some overall lessons I learnt

  • If you want lots of citations write open-source software and get feedback from users
  • Get users or non-expert collaborators to help write the instruction manual
  • Keep (at least the first paper) as simple as possible
  • Use social media and answer queries as quickly as you can

Part 3: 5 pieces of advice

1. You can’t
make your
work
completely
reproducible

2. … but try to make every piece of work you do as reproducible as possible

  1. Make it open access (pretty much compulsory these days)
  2. Make it so that people can re-create your graphs
  3. Make it so that they can generalise your code

3. Aim to turn every paper you write into an R package

4. Write better code and comment it properly

5. Don’t walk away from your hard work if you want to reap the benefits from it

Some useful resources

The future!

  1. All journals requiring open-source code/data that runs and reproduces the figures
  2. The ability to submit data (done) and the ability to submit code and get a DOI
  3. Automated reproducibility

Summary

  1. Making your papers/code reproducible is to your advantage* (*if you’re intending to pursue a career in academia)
  2. Learn the tools of the trade (e.g. R packages, Rmarkdown, GitHub) for reproducibility
  3. Make reproducibility a cornerstone of every paper you write
  4. Do you best and don’t beat yourself up about not making everything reproducible

Back to the IPCC results

Resources and funding