Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Selfish reasons to carry out reproducible research

96e8ca061c005a42d360459d366ec923?s=47 Dave Lunt
January 15, 2021

Selfish reasons to carry out reproducible research

Given 29 November 2019 at the Univerisity of Hull

Inspired by Markowetz F. Five selfish reasons to work reproducibly. Genome Biol. 2015;16: 274. doi:10.1186/s13059-015-0850-7

These Google slides are at https://bit.ly/35yDvIG
The slides created by me (almost all) can be considered CC0 public domain, use as you will. Some parts were created by others, I have put credit and copyright info in the speaker notes to the Google slides


Dave Lunt

January 15, 2021

More Decks by Dave Lunt

Other Decks in Research


  1. Selfish reasons to carry out reproducible research Dave Lunt dave.lunt@gmail.com

    @davelunt https:/ /bit.ly/35yDvIG
  2. What is reproducibility?

  3. first Why, then How

  4. It's required Why be reproducible?

  5. McNutt M. Journals unite for reproducibility. Science. 2014;346: 679. doi:10.1126/science.aaa1724

  6. None
  7. None
  8. RCUK – Statement of Expectations for Postgraduate Training Students should

    receive training in experimental design and statistics appropriate to their disciplines, and in the importance of ensuring research results are robust and reproducible
  9. None
  10. None
  11. It's the right thing to do It's science

  12. Ask not what you can do for reproducibility, but what

    reproducibility can do for you Florian Markowetz
  13. It will save you time and effort It will advance

    your career Selfish Reproducible Research
  14. Who here has tried to reproduce a published analysis? Who

    is most likely to reproduce your work?
  15. Do experiments work first time for you?

  16. “Future You” will be most likely person to reproduce your

  17. Future You Previous You Previous You does not respond to

  18. It will greatly help “future you” Selfish reasons to carry

    out reproducible research
  19. How can we save time, effort? eg: make figures from

    scripts this is reproducible analysis
  20. Your research will be faster and easier (and better)

  21. The old way

  22. None
  23. None
  24. Automated reproducible Manual Cumulative total effort Number of repeats Yes

    you will cross this point
  25. Reproducibility makes it easier to write papers and respond to

  26. Reproducible research will save you time and effort Reuse and

    recycle data generation and analysis
  27. Errors are ubiquitous Retractions will hurt you Reproducibility helps your

  28. Reproducibility will help your career Reputation Rigour New collaborators Rapid

    Agile Future-proof
  29. Choose a collaborator Rigorous, modern, open, with future-proof methods. Leading

    the way. Prepared and shared many of the methods you need already.
  30. Projects are not unique. How will you build your career?

  31. Required Helps “Future You” Easier & faster, agile Easier papers

    Helps your next project Builds your career Avoid major screw-ups Makes you a cool collaborator Selfish reasons to be reproducible
  32. Pause But what about ...?

  33. I’d rather do real science than tidy my data It's

    the way I’ve always done things, and I’ve got this far Excel is just fine My data and code are spread across many computers, I couldn’t do this I’ll sort this out at the end My field is too competitive, I can’t slow down to do this
  34. I’m not a computational biologist

  35. How?

  36. 1. relax, most problems are solved

  37. 2. think of it as training

  38. 3. celebrate the quick wins

  39. Quick win: Be part of a support community

  40. Make 1 figure from a script Quick win

  41. Butterfly_project - DATA -raw_data -fig1_data - FIGURES -fig1.pdf -fig2.pdf -table1.md

    - RESULTS -PCA -lin_regr - SCRIPTS -fig1.py - README.txt Informative names Structured Text description of what is where Spend 1 morning to organise your data Quick win => Provenance and persistence
  42. 2. think of it as training

  43. Wilkinson et al. The FAIR Guiding Principles for scientific data

    management and stewardship. Sci Data. 2016;3: 160018. doi:10.1038/sdata.2016.18 Records, Coding, Workflows, & Research Objects
  44. Make data open with a doi Findable Accessible Interoperable Reusable

    Quick win zenodo.org figshare.com osf.io For you (and for others) Yes, data can be private until you’re ready
  45. zenodo.org Yes you can keep data private until publication

  46. It's free osf.io

  47. osf.io

  48. None
  49. File storage Integration of GDrive, Box, Dropbox, Git etc OSF

    cloud storage Everything in one place
  50. Activity All changes recorded with version control Roll back to

    previous versions Comments and collaborations
  51. Components are folders Structure and backup Robust sharing and privacy

    Can be published with doi
  52. try osf.io Easy to organise project Easy to store &

    publish data Easy to collaborate Easy reproducibility
  53. Making labwork reproducible protocols.io

  54. It's free

  55. Quick win METHODS SECTION Experimental procedures are briefly described here

    for context, and exact protocols and reagents are detailed in doi:1234567 and doi:987654
  56. Summary

  57. It will save you time & effort Selfish reasons to

    be reproducible Write once and iterate, faster, helps with ms, helps with reviewers, don’t start projects from scratch- build on prior reproducibility
  58. It will advance your career Selfish reasons to be reproducible

    Fast, cutting edge, future-proof, you’ll look good, more collaborators, extra citations, avoid career-ending disasters, builds a group etc etc
  59. Do not try to be completely reproducible! Shocking finale PTO...

  60. Do not decide to be reproducible. Decide to be a

    bit more reproducible, celebrate the small wins. Spread the word. Take home message
  61. None