Day 21

Announcements

  • Presentations will be 5 minutes per group on Thursday, November 12 (i.e., 1 week from today)
  • Draft reports (~6 pages) due on Thursday, November 12 at 21:00
  • Quiz on Monday, November 9 will cover hw2 material, scope, plotting, basic linear modeling as covered in Matthew’s lectures, basic Git workflow, and possibly some general questions about hypothesis testing.
  • BIC tour on Monday, November 23 (during lab time)

Reports and presentations

“Seek simplicity, and distrust it.” — Alfred North Whitehead

  1. Briefly describe data (~30s)
    • Paper
    • dsnum from openfmri.org
    • What kind of data is it?
  2. Briefly describe what you’ve done so far (~30s)
    • data fetching/preprocessing
    • initial analysis
    • plots? figures?
  3. What is your plan? (~3m)
    • What analysis will you perform?
    • How does it relate to the original data analysis?
      • Will you use all the data? Why or why not?
      • What model/preprocessing steps will you simplify?
    • What are the problems you face?
      • Try to be explicit about your issues?
      • Suggest potential solutions and/or approaches.
      • What do you need to research more? Have you found sources?
      • Will you try to make “inferences” from the data?
        • How will you deal with multiple comparisons?
        • How will you attempt to validate your model?
  4. Process (~1m)
    • What was the hardest part of the process so far?
      • Git workflow, Python, fMRI data, all of the above
      • Having an ill-defined assignment?
    • How successful have you been at overcoming these obstacles?
    • What issues have you faced working as a team? How have you been addressing them?
    • What parts of the class have been the most useful?
    • What parts have been the least helpful or most confusing?
    • What do you need to do to successfully complete the project?
    • How difficult are you finding it to make your work reproducible? Do you feel confident that you are in a position to make your projects reproducible?
    • What would be most helpful for your team in the remaining weeks? Additional lectures? Structured or unstructured group work?
    • What potential topics would be most useful for us to cover?
      • Overview of brain / neuroanatomy?
      • More linear regression (ANOVA)? PCA? The mathematics or the implementation?
      • Machine learning (classification, prediction, cross-validation)?
      • Permutation tests (and maybe bootstrap)?
      • Software tools (Git, Make, Python, statmodels, etc.)
      • Technical writing and scientific visualization?
      • Advanced topics (regularized regression, selective inference)?

Remember

  • If you have questions about your projects, there should be a pull request with code or text and use @jarrodmillman, @matthew-brett, @rossbar, andor @jbpoline
  • Most of your code should be written as a collection of functions with tests, then use scripts calling these functions to perform your analysis