The next dissertation steps
Well it’s the weekend! Which means work, because of course it does. If it weren’t for the fact that I’m doing both work-work (as in working in a hospital) and PhD work (school), I would have my weekends (mostly) free for myself, but since I want to graduate soon, there’s a lot that needs to happen. Order of operations matter and today I’m going to talk about my plan to get the first phase of my dissertation data finalized. And it needs to happen quickly!
DAPRA is coming! Well the DARPA forward conference that is. With just a few weeks to the deadline to submit my poster and slides for the talk (if I get selected, fingers crossed on that front), I have so, so, much work to do between now and then! Now, I need to figure out the juggling act that needs to happen. On one hand, I could do an almost complete (or mostly complete) analysis of a single dataset. On the other, I could completely process all the data and do a basic analysis of the whole dataset. I know which one I want to do and today I’m going to justify that choice (even if it’s just to myself).
For me, the answer is obvious. I’m going to finish the preprocessing and aligning steps for all (or at least “enough”) of the data I’ve collected. I’ve finished exactly 1 dataset so I only have 9 more to go… woo. While that sounds like a lot, most of it is semi-automated. I tend to be hands on when it comes to processing data because that’s the most important step in my opinion, so I don’t like to fully automate anything since each dataset is unique enough that there are subtle differences that need to be accounted for. Certain filter parameters that need to be adjusted so you don’t overclean (remove useful data) or underclean (leave noise or unwanted signal) your data.
While having a complete analysis done on a single dataset would be impressive, it’s not enough for me. I want to have some group data. This goes back to the entire years worth of me talking about super secret technique (SST) and having a rather huge limitation on what I did (n = 1). I want to have a group analysis to show that this works or I will end up with a whole lot of nothing (in my opinion) at the end of the day. I don’t want to be relying on any n = 1 dataset since I have far more than I need to find statistical significance and show that this is generalizable.
With the DARPA event coming I would feel more comfortable showing that it works, but only have scratched the surface, than to show something amazing with the giant asterisks on it that this was a single dataset. And who knows, maybe I can do a bit of both, complete the basic analysis for the full dataset, then go back and do a very detailed analysis for a single participant. Frankly if that happens then it’s more of a case where it’s just as easy to do the analysis for the entire dataset as a group, but you never know, maybe I’ll run into problems.
All this hinges on the DARPA event because that’s the next major (and really only) milestone I need to hit. Every other deadline I have is self imposed and the only “real” deadline after this is my graduation deadline. Which of course, I’m anxious to hit and only have a few months to get a big chunk of the work done for that to happen, but we’ll make it… I hope. It is a lot of work after all.
I’m hoping this weekend to make significant progress in the preprocessing for all the datasets I have. If I can get even just a few done, I’ll be very happy. My goal is to hit at least half of them before the deadline so I have time to do analysis of the data. But if I get more I’ll be okay with that, I really don’t want to hit less, but that could (will probably) happen as well. I’m going to go easy on myself in that regard though. Next weekend will be the only real time I have to do data analysis so we’ll know very soon how many datasets I can process in a day or two.
Fingers crossed! While this part isn’t as physical, it’s just as much (if not more) work than it took to collect the data. I just hope everything plays nicely, because if it doesn’t, well I really don’t want to have to think about that. Right now, it’s go time.
I think you’re making a good decision. An interesting n=1 result was something you already had, so getting another one of those feels less compelling than finishing the shallow group analysis. I hope you had good success with the pre-processing today!
LikeLiked by 1 person
October 10, 2022 at 12:46 am
Thank you (and welcome back!) I’m glad you agree since I really value your opinion and insights. It’s been a long couple of days, but I’ve made a lot of progress!
LikeLiked by 1 person
October 10, 2022 at 9:27 pm