Can the practices of the Quantified Self community be characterised as using the Scientific Method, or should it be distinguished as something else? In particular, does performing research one oneself (n=1) disqualify the use of the term as it has to be applied on n=N and generate general theories?
Quantified Self experiments can be scientific (and have been published in scientific journals), but are not clinical research due to the n=1.
The same is true for clinical practice vs clinical research, though “precision medicine” might blur the boundaries.
Yes. It’s science. It may not be as strong with n=1 (weak experimental method) but it’s the scientific method alright.
gathering evidence to inform better models.
Thanks for the answer, Eric!
Do you have any links to these published articles?
When thinking about how to communicate this distinction to others, this is what I’ve come up with:
- Using the Scientific Method with n=1 is done with the purpose to apply current knowledge to the individual. E.g. the Scientific Method is used in surgery on an individual: finding medical issues through observation, questioning how they can be fixed, hypothesising through pulling evidence based research on the human body, predicting how a specific surgical procedure should help the individual, testing/performing the surgery, producing results, and finally analysing these results. The Scientific Method has been used but the outcome is not necessarily general theories, but a problem solved for the individual.
- Using the Scientific Method with n=N (large number) is done with the purpose to test hypotheses that are general statements about nature, and has to be done in a statistically significant way to ensure that it was not other factors or random noise that produced the results. E.g. an initial hypothesis of F=ma in physics needs to be tested in a controlled, statistical significant manner to support this general statement of force, mass and acceleration in nature.
Does this seem as a fair explanation? Are there any flaws in the reasoning or is there a better way to communicate these ideas in your opinion?
Thanks for the link, Elo!
See also this call for papers, don’t know when/if this special issue will be published.
Yes, but sometimes it can go beyond “current knowledge”, and come up with new ideas to evaluate on larger numbers of people.
I am in the midst of publishing a n=1 study for myself. PM and I can send you a preprint of the paper. My co-author is a GI prof at hopkins.
The scientific method involves testing hypotheses. There is no core requirement of “n=N”, or even need for general applicability of the outcome.
I think Gustavo’s point is important. Why would the number of subjects be a key factor in judging whether a bit of science is sound except when claims are made about groups and the method of proving these claims involves sampling?
@Elo I’m not sure I’d call it a “weak” experimental method. As we head towards precision medicine, there are definitely lapses in the one-size-fits-all approach, and n=1 begins to address that. We’re probably still a considerable way from implementing all that, but well…
As a side-note, a fascinating piece of fiction around all this is the recent film Realive
Talking about the experimental method now not the scientific method. With N=1 you can still have 100 trials (or a high number of trials) to create validity.
Assuming a valid, accurate experiment you can show that “this happened”, (when I don’t sleep I feel bad the next day). The extent to which that evidence can be applied to other people is what can be called into question. And yes.
- There is limited information in a small study.
- It may be hard to say how applicable the results are
- it may be hard to confirm validity of the results due to problems with confounding variables.
And more. There are problems. But there is still information value in n=1 studies.
A bit confused now. If a 100 participants (or did you mean trials) are used to show “this happened”, then isn’t it already n=100? Doesn’t that stop being n=1?
one participant (n=1) over 100 days. 100 trials, same person but one each day.
Compared to (n=1000) participants, with 3 trials each.
Yes, language is vague in trying to explain this.
If you are testing the question, “can any person live without their head attached to their body?” you only need one trial to succeed in order to show that, “yes, one guy could”. However it’s harder to show that “no one could live without their head attached to their body” unless we test for each person. (or somehow know something about physiology). Or run a representative number of trials (say a million people to be conclusive) Still it only takes on to disprove the question. So there can be information value in one trial.
I’m not sure I fully understand or agree! 100 trials can’t be the same person over 100 days, that’s still one trial. And that method could just confirm the same fact/bias repeatedly, in a loop, not validate it by other means. Guess this needs further thought, and you’re right, the jargon/language unfortunately doesn’t help!
What about these examples of n=1 experiments: The Voyager mission to Mars, tests conducted before/after the Mount St. Helens volcanic eruption, analysis of the effect of fiscal and monetary policy on the Great Depression. Lots of science is conducted on n=1 situations.
Or run a representative number of trials (say a million people to be conclusive)
Whoa, that got dark fast
Isaac, i’m also interested in your n-of-1 paper. (I don’t have a PM option for some reason).
Hi Isaac, I’d love to have a look at this study but cannot send you a PM for some reason, similar to pkuhar.
Could you send me a PM?