As Louie likes to say, we’re now really turning the crank on our 2-3 main experiments!! And we even got a bonus stats/R crash course, too. Here’s a summary of all of that and more:
Data Analysis Review:
- Louie cleaned up the time-lapse trial data to make sure it was error-free and easy to read.
- We also discussed the unique nature of the time-lapse data – note that our variables are mostly categorical, rather than numerical
- One way to analyze this is by using a GTest, which is essentially a fancy updated chi-squared test; this kind of analysis compares observed vs. expected values, to see how far they differ from 50-50 (random).
- Remember to install DescTools in order to use GTest! These are the steps: Tools > install packages > DescTools > install DescTools
- There are lots of ways to skin a dataset… so try different things, like treating proportion stripes as a continuous variable & using the lm function.
- Even a “small” dataset can be looked at in many cool, complex ways.
- We considered whether the flies’ “decisions” were independent of one another – keep this in mind as you choose your data analysis method! There are rarely correct answer in statistics, just ways to justify your assumptions.
- Louie showed us some of his data plots, which had some interesting results… orientation seems to have some weird effect on proportion of time spent on stripes. We’ll find out after more trials if this is a fluke or if it has some other cause (possibly temp or light from the window?).
- Note: the preliminary error bars are fairly substantial, so “don’t drink the koolaid yet“! But the deviation from 0.5 is pretty exciting…
After going FAR down the stats-rabbit-hole, we split into our 3 teams: Team Time-Lapse, Team Flypaper, and Team Choice-Experiment. Team Time-Lapse made a lot of progress with scoring pictures, and also did some trials outside. Team Flypaper wrangled with more sticky substances, as usual, and come up with 2 final model designs. Time Choice-Experiment finalized their arena design as well, and analyzed some of that data.
As usual, the flies LOVED their new mesh-topped homes, so team Time-lapse will likely have to consider this in their picture scoring. Team Flypaper dealt with the challenges of various materials and settled on a tube model and a 2D model, which they hope to set out on Tuesday – or whenever the rain clears up.
Finally, Team Choice reported back that there seemed to be no significant difference between # of flies that chose stripes vs. non-stripes. Remember that “no significant difference” doesn’t mean no result! It would be really interesting if their arena yielded different results than the time-lapse trials.
Whew! We kicked butt this week – great job everyone for being so productive, creative, and overall awesome. Here’s my plan for Tuesday:
- For the most part, let’s continue splitting into our respective groups.
- It seems like Team Time-Lapse needs some extra hands (particularly for the “magic dark arts” of photo-uploading), and that Team Choice is winding down.
- Remember to be thinking about what stats you’d like to analyze, and how you want to write-up the experiments we’ve done so far in the final paper.
- Team Flypaper will hopefully finish making the 2D models, and set them out (either at EE or the channel) along with the tube models!
Have a great weekend everyone! 🙂