Recording Context and Confounding Factors

What context and confounders are

Your numbers do not exist alone. A poor night's sleep may follow a late meal, a stressful day, a noisy room, an illness or a time-zone change. In research, a confounder is a variable that can make an apparent effect look larger, smaller or different from the true one. Researchers explain that to be a potential confounder a variable must be associated with the outcome, must be unequally distributed between the groups being compared, and must not be an effect of the exposure itself. [1]

In a personal experiment, your "groups" are the days or weeks with and without your change. If those days differ in other ways, the comparison is unfair. This guide is information only and applies to ordinary habits; it is not a program.

Why it matters for one person

A formal single-person study compares periods under an intervention and a comparator, and the investigators in one published example collected not only mood but also alcohol units and social interactions every eight hours, so that other causes of mood could be adjusted for in the analysis. [2] You do not need that statistical machinery. You can still note the same kinds of things so that you can judge a result honestly.

What to jot down beside your main outcome

Keep it short, or you will stop. Examples:

• Sleep and routine: unusual bedtime, late meal, alcohol, caffeine, screens late at night, noise, a different bed.
• Health: illness, pain, menstrual cycle if relevant, new medicines or changes in dose (taken as prescribed; this is a note, not a suggestion).
• Activity and stress: an unusually hard day, a stressful event, travel, a long drive.
• Environment: heat, seasons, daylight. Daily resting heart rate, for instance, showed a small but significant seasonal trend in one large study. [3]
• Measurement: whether the device was worn, charged and fitted the same way.
• What you expect: a line about whether you hope the change will help.

Reading the notes

• Look at unusual days first. If a flagged day is driving your "result," be cautious about the conclusion.
• Do not delete bad days. Keep them, and mark them.
• Do not overinterpret. Adding many notes tempts you to explain every wiggle. The aim is to spot obvious rival explanations, not to find a story.
• Know that you cannot catch everything. Unrecorded factors remain. A comparison that cannot rule them out is an observation, not proof. See correlation vs causation in personal data.

Expectation is also context

How you feel is shaped partly by what you believe. NCCIH describes the placebo effect as a beneficial health outcome resulting from a person's anticipation that an intervention will help. [4] In studies of perceived sleep, participants told they had spent more or less time in REM sleep performed differently on cognitive tasks, depending on what they were told rather than how they reported sleeping. [5] Notes about expectation, and rating your outcome before you look at any tracker data, can help. See placebo and expectancy effects in wellness tracking.

When context points to a health problem

If your notes keep showing symptoms such as persistent exhaustion, loud snoring with pauses in breathing, chest discomfort or fainting, stop experimenting and speak to a clinician. See talking with a health professional.