L-Tyrosine has a "B" rating on Examine.com for its potential effects on attention, mentioning that studies suggest it may be able to preserve attention under conditions of stress, but evidence is limited beyond that.
I am currently in the middle of a blinded trial with L-Tyrosine. Over the course of 20 days during which I intend to focus for sustained periods, I take either 500mg L-Tyrosine or a placebo and record my focus on a scale of 0-10.
This market will be resolved YES if the mean for focus on days taking L-Tyrosine is higher than days taking the placebo.
(This is my first market so I'll keep the resolution criteria simple, but in future I'd like to use a simple model to estimate the posterior distribution of % difference between the placebo and supplement using PyMC, and then resolve YES if that distribution suggests a >50% chance of 2.5% increase in focus/whatever I'm trying to measure. Anyone can feel free to comment if they think that's worth doing or if there's a better approach!)
Extra notes:
- I didn't pick L-Tyrosine for any reason other than it's mentioned on Examine.com, and it's easy to get a hold of.
- I'm 10 days into the trial, but I won't mention how it's gone so far until the trial is complete.
- I won't be betting on this market.
1,000
3.00Does anyone have a preference for how my next N-of-1 experiment market should be resolved? If I again used a naive model and said "This resolves YES if the model suggests a 75%+ chance that there is any improvement/reduction in {effect}", would that be too convoluted?
For reference, after 20 data points and a difference of 7.0 (supplement) vs. 7.4 (placebo), the model suggested 19.4% chance of any improvement to focus, and conversely an 80.6% chance it's actually harmful.
I could also pick a point on the distribution like "Resolves YES if there is a 50%+ chance there is at least a 2.5% improvement/reduction in {effect}", as per the image below (L-Tyrosine not a great example, but you get the gist):
Mean focus with L-Tyrosine was 7.0 while mean focus with the placebo was 7.4, so this question has been resolved NO. I also created a basic model to estimate the posterior distribution of the % difference in focus between the two, and it returned a 19.4% chance that L-Tyrosine has any positive effect on my focus.
@LuisCostigan damn. Oh well, stand by my trades that this was effectively a coin flip. Not sure why it was consistently below 50.
@Eliza Haha, nice work! (Disclaimer: I developed the two apps mentioned below)
Yes, not set in stone but I'd like to setup an experiment with L-Theanine to see if it reduces my anxiety when I consume caffeine: https://n1.tools/experiments/anxiety/lTheanine
I'd also like someone to setup a market for Binaural Beats for focus, using placebobeats.com to generate real and placebo beats. I'd do it myself, but I feel like there's a chance I'd recognise the tones.
This is just me perhaps, but I think it would be cool for people to run N-of-1 experiments, generate their own data, and incorporate prediction markets as part of the process. Maybe the prediction market would provide the prior likelihood, and that can be updated with user data from their own experiment (although I'm still considering exactly what that looks like).
@MaxTroet Interesting to hear your experience. Yeah I haven't been able to tell so far, I take it blinded with a swig of Huel and the thickness hides any difference. If I somehow found out, I simply wouldn't record the data point.
@LachlanMunro No statistical test at all, I kept the resolution criteria very simple. Your understanding is correct.