File:Nootropic Survey 2016 Factor Analysis.png

An analysis of data from the 2016 Nootropics Survey by gwern, using factor analysis to determine correlations between preferences between different nootropics.

Original source.

Analysis by gwern

What do these factors mean?

g is the general factor and works about as expected: those who rate one nootropic highly tend to rate others highly as well, and this explains ~0.26 of the total variance or ~R^2=50.1%, I think.

F1 is kinda a bit like anxiolytic/adaptogen, F2 is maybe stimulants (interesting to see LSD microdosing lumped in with noopept/aniracetam/oxiracetam – having difficulty sleeping after an LSD trip is commonly reported due to the stimulating effects), F7 looks like another stimulant cluster, F9 looks like it’s just for tea/coffee users, F3 I think must be a -racetam cluster (noopept/piracetam/choline/inositol/DMAE, no idea about theanine/Ashwagandha)…

But F4/5/6/8? I have no idea. F4 I would call a stimulant cluster except that along with the Semax/LSD/Nicotine loadings it also pulls in Picamilon & Selank, which I thought were supposed to be more anxiolytics! F5 might be more of a anxiolytic cluster but I don’t know enough about Uridine/PRL853 to know where those should go. And F8 is weird for mostly having negative loadings (-0.85 for Adderall! this is the only factor which includes Adderall too. huh?)

(You might think that retrying with `nfactors=2` might at least reproduce the usual stimulant vs anxiolytic split, but nope. The two non-g factors split up the stims about evenly: F1 gets Modafinil/Caffeine/Coluracetam/DMAE/Theanine/Noopept/Oxiracetam/Aniracetam/Rhodiola/Creatine/Ashwagandha/Bacopa/Choline while F2 gets Phenylpiracetam/SemaxandNASemaxetc/Melatonin/Tianeptine/LSD/Adderall-/Phenibut/Nicotine//Uridine/Picamilon/SelankandNASelanketc/PRL853. Very weird. Do we need to resort to deeper classifications like neurotransmitters? I’m not very expert on that.)