Research Update

1/13/23

 

Stephen

Paper Revision

  • This is the main task I am working on
  • Received initial review back on 11/21/22
  • Revisions "must be submitted within 3 months"

Reviewer 1

  • Main concern is that the paper is of 'limited generalizability'
    • Only uses one dataset. How would it generalize to other stimuli or tasks in RSA studies?
  • Not clear what the contribution is
    • Increasing N improves signal. Need to clarify what the results add
  • Some citation questions
  • Following points need to be addressed:
  • a) can adress other stimuli
    • include other datasets
  • b) 6 presentations is a lot. In block designs this would not be, but we could get at this by considering the 'conditions' to be the categories
    • In SOE data, 8 categories each with 6*40 repetitions
  • c) can be fixed by doing analysis in different ROI's, parcellation
  • d) seems like wording problem

Reviewer 2

  • Thinks we should spend more time discussing the generalizability of the results

Reviewer 2

  • Thinks we should spend more time discussing the generalizability of the results
    • elaborate on what conditions we think the results should hold for
  • Encouraged to explore in more depth how improvements in reliability change discriminability of the models
  • Clarify formula
  • Reference a double bootstrap paper
  • Expand on relationship between group average RDM sensitivity and statistical inference

Thoughts

  • Reviewer 1 is harder to address
  • To show generalizability, wanted to show other datasets
    • Show CAT data
      • Adds 40 participants
    • Show ERV data
      • Shows generalizability to different kinds of words, and we actually changed experimental design
    • Show AVA data
  • While it is all same task, different participants, studies, words

Re-processing Data

  • Since the original submission of the paper, I have reprocessed all data through a new version of fmriprep
    • Don't even have the old data, plus I wanted to add in CAT data
  • Spent some time trying to figure out how to exclude participants
    • Calculate correlation between mean FD and mean RSA correlation
    • Calculate correlation between mean Fstat and mean RSA correlation
    • Looked at within participant reliability and average RSA
    • Created a new measure of consistency
  • No easy way to exclude ...
  • Can deal with later 
  • BUT
Made with Slides.com