Quality Control Checks

[TomDonoghue]

We want to check our results (and the paired task results) with relation to quality, for example:

• are significant cells representative in terms of firing rate and general quality
  • e.g.: it's not just very low (or high firing) cells that are significant
• are significant cells well powered and robust across trials
  • e.g.: some sessions have fewer trials - do these contribute a bunch of cells? Should they be dropped?
  • double checking consistency (not outlier trials driving things)
• are significant cells found across sessions & subjects
  • e.g.: it's not just some weird / specific session or subject contributing all the cells
  • Note: this also relates to comparing to behaviour (maybe bad behaviour sessions are not representative / have no cells - could drop)

Potential ToDos:

• drop cells from our analyses
  • note: we currently only drop based on 1-back quality labels
• drop session from our analyses
  • sessions with too few trials / too bad behaviour