From Start to Finish: What Goes Into a Study -- Day 235: Prelim preprocessing

This series of posts is documenting the journey of my first first-author project, from the infancy of the research through publishing. I am highlighting the major checkpoints of the project — when things move forward or backward — rather than a daily update because that would seriously be boring. Just about all the content discussed will be directly related to the project I’m working on.

If you haven’t read the previous posts, check these out!
Day 1
Day 2 – 7
Day 27
Day 82
Day 145
Day 193
Day 197
Day 203
Day 216
Day 224
Day 233

Check out the next day… Day 299!

Day 235: Preliminary data preprocessing

In the land of neuroimaging, recorded data goes through a step called "preprocessing" is generally needed. A lot of data recorded using EEG in our project is not used in the analysis. In our project, we have 6-second windows for which we assume brain activation will be occurring. Anything outside of that window (depending on the event) is not needed. We have four events: a resting baseline event, a prechoice event, a postchoice event and an outcome reaction event. That's 24 seconds per choice and there are 24 choices total, making it only about 10 minutes of data per participant. That's really only 25% - 33% of the total recording time (they are usually in the experiment for 30 - 40 mins). The other 66% - 75% is unnecessary.

So we preprocess, which involves separating out the data we are actually going to analyze from the data we aren't going to analyze. But we do more than just crop data.

EEG is sensitive to things like muscle tension, eye blinks, teeth clenching, saliva swallowing, other electrical deshielding... all sorts of things. Luckily, the community of EEG preprocessors have figured out how to accurately remove these reoccurring ("stereotypical") artifacts. There's a whole process of filtering and separating out events and removing artifacts that I go through until the data is absent of artifacts. Generally, a nice dataset has 5% of the usable data rejected, leaving us with 95% of about 10 minutes of data per participant.

At this point, we will have the data we want in the way we want it!

But that's just getting the data. There's a whole battery of statistical data analyses that we'll have to go through as well. I'm hoping to have all preprocessing done by April 7th, with stats analyses coming between April 7th and April 21st. Then, from April 21st to May 8th, I'll be writing the abstract with my PI, which we'll be submitting to Society for Neuroscience!

Might not have many updates before prelim is done and stats begin, so stay tuned!

Check out Day 299 here!

Leave a Reply