Tomorrow I’m going to start an experiment where I test eleven different interventions on variable schedules for the next 100 days, measuring my sleep and my cognitive performance each day. Here are the interventions I want to test, along with the number of days before the interval repeats.
Posture 3 (sitting/standing/walking during the test)
Music 4 (listening/not listening during the test)
Krill oil 6
Acetyl-L Carnitine 26
Whey Protein 34
Gluten 46 (can’t just test positive hypotheses ;))
So I would eat butter for 5 days, then not eat it for 5 days, etc. Each condition is exposed to every other condition as equally as simply possible, and to any other trends over time not measured. Since I expect most of the interventions to have no effect, I maximize the experimental efficiency by running them all at once.
Sleep I’m going to measure with the Zeo. I have some other subjective measurements, too. But the main point is to measure cognitive performance. I’m going to spend some time each day taking cognitive tests, trying to suss out any effects from the interventions.
So my questions are, to those of you with some experience with the interventions listed: how long after consuming each substance should it have the maximum effects? I was thinking of eating them at noon, and then measuring maybe at 3pm and maybe again at night, but I don’t know much about most of them.
Any other feedback on how to design the experiment is welcome, too.
too complex by far. Suppose one day is unusual for reasons that have nothing to do with all these variables…I think you will have a hard time figuring that out. Before starting such an ambitious experiment, I would test much simpler versions.
Thanks for the critique. I am not sure why an unusual day would gum things up, though. My analysis plan is to look at each intervention separately, and for each type of test/metric that I collect, split all the results into two groups: with intervention, and without. So if I do 20 working memory trials a day and 50 simple reaction time trials, etc., then I would look at the 1000 working memory trials with creatine and see if they’re statistically significantly different from the 1000 without creatine, and the same for the 2500/2500 reaction time trials, and the other tests, along with the 50/50 nights of ZQ scores, etc. Then for each pair that showed a significant effect, I would repeat the experiment on those interventions later to confirm. It seems straightforward to me–am I missing something? I am not a statistician.
Oh yeah, and if I got better at the tests over time, I would try to apply an adjustment so that the later trials had the same average as the earlier trials. I’ve already been taking the tests for a few weeks and they’re very simple, so it might not be a big issue.
Thanks Bossman. The creatine will be 19 days on, 19 days off. Maybe not long enough for guaranteed loading / washout, but for a preliminary investigation, perhaps it’s long enough? I’ll drink it with orange juice–thanks for the tip.
Looks like the study used 5g doses, at “the same time each day”. I’ll try that, in the mornings.