Glossary

01A to C

AB design

The simplest one-person test: a stretch of days before a change (A), then a stretch after it (B). With no repeat, it can’t rule out other explanations for a difference. See Design the test.

Abstract

The short summary at the top of a research paper. Many papers can be read only as an abstract without paying, so the References page says when a summary rests on the abstract alone.

Baseline

A stretch of ordinary days, recorded the same way as the test, that the test is compared against. See Baseline first.

Blinded

Kept from knowing which period is which. In a blinded study, the person (and often the researchers) doesn’t know whether they’re having the real thing or a dummy. See Expectation.

Carryover

When the effect of one period spills into the next, so the second period isn’t a clean comparison. A gap between periods, called a washout, is meant to stop it.

Confounder

Something else that changed at the same time as the thing being tested, such as poor sleep or a stressful week, and could explain a difference instead. See Record what else was going on.

Crossover

A design where the same person has each condition in turn, rather than different people having different conditions.

02D to M

DOI

A digital object identifier: a permanent link to a published paper, written like 10.1136/bmj.324.7347.1193. The guide links to each paper’s DOI where it has one.

Double-blind

Blinded for both the person in the study and the researchers who see them.

Ecological momentary assessment

Rating how you feel at the moment, several times a day, in everyday life, rather than from memory later. Also called experience sampling. See Write it down on the day.

et al.

Short for “and others”: “Stone et al.” means a paper by Stone and co-authors.

Food challenge

A test run under medical supervision, in which a food is given in measured, rising amounts while trained staff watch. Bringing a food back at home is not a food challenge: see Safety, and when to involve a clinician.

Main measure

The one rating, chosen before a test starts, that the result will be judged on. Research reports call it the primary outcome. See Choose what to measure.

03N to R

n-of-1 trial

A trial in one person, who goes through repeated periods with and without the thing being tested, usually in a random order and often blinded. “n” is the number of people in a study.

Nocebo

The opposite of placebo: unpleasant symptoms that come from expecting harm, not from the thing itself. See Expectation.

Open test

A test where you know what you changed and when. Most self-experiments are open, which leaves room for expectation to shape the result.

Placebo

A dummy, such as a pill or powder with nothing active in it, given so that people in a study can’t tell which period is which.

PubMed

A free public index of medical research papers, run by the US National Library of Medicine. Each paper has a PubMed number.

Randomised

Put in an order set by chance, such as a coin toss, rather than chosen. A random order makes it unlikely that the timing lines up with good or bad days, though by chance it sometimes can.

Recall

Remembering how you felt over a past stretch of time, rather than noting it at the time. Recall over a week tends to run high.

Regression to the mean

The tendency of an unusually bad or good spell to be followed by a more ordinary one, on its own. Starting a change at a bad patch can make it look as if the change helped. See Baseline first.

Reintroduction

Bringing a food back after leaving it out, one food at a time, while watching for changes.

04S to Z

Single-case design

A research approach from education and behavioural science that studies one person (or a few) over time, with repeated phases, read mainly from a chart.

Systematic review

A review that searches for every study on a question in a set, planned way, and sums up what they found together.

Threshold

The size of change, decided before a test, that you’ll count as big enough to matter. A change past it can still be an ordinary swing, which is why tests repeat.

Washout

A gap between test periods, long enough for the effect of one period to fade before the next begins.