Quantified Self and n1.tools

I met the creator of n1.tools at a conference recently. He’s great and n1.tools is great and I have a lot more to say about n-of-1 experiments (we did an elaborate one with our son recently). For now I just wanted to start the topic here and see who else is game to talk about self-experiments and what tools you use.

Teaser screenshot:



I had not heard of n1.tools but that is super cool and I’m definitely going to give it a try. I’m just trying to decide what experiment to try first!


Does it do a good job of explaining graphs like that to anyone who doesn’t have a math or statistics background?


Oh, sorry, I should’ve clarified that that teaser screenshot is my own from Mathematica, not connected to n1.tools, and it’s super not self-explanatory.

(Quick stab at explaining it: The different treatments are A, B, C, and D, one of which was a placebo. Being measured is a number from 1 to 5 quantifying an outcome the treatments were hoping to affect. The top graph shows cumulative distributions based on the n=12 samples in each treatment. The sample means (μ) and 95% confidence intervals on those means are shown on the right. At the bottom is a visual depiction of the 95% confidence intervals. So D (red) is pretty significantly better than C (blue), for example, since the confidence intervals don’t overlap.)


I want to use it, but what I remember of the underlying model was that I’d have to shoehorn my experiments in.

I have a running list of experiments I’d like to run generally, but some have hour-by-hour or minute-by-minute effects, some could be tested daily but the models I want to evaluate have hidden confounders and/or multi-day effects, etc. My brain may just be hopelessly poisoned by WebPPL’s flexibility. :sweat_smile:

So I haven’t found a use for it yet. I intend to try it out as soon as I do, though!