Psychology & Behavioral Sciences: Plot my own data

What you'll be able to do: Turn a few weeks of your own sleep, mood or reaction-time numbers into a chart, and see effect size, noise and correlation-versus-causation stop being exam vocabulary and start being things your own data actually does.

Scope note. The natural sciences' version (article 09 there) plots practical results against a physical model, a titration curve, a spring's extension, where the expected relationship comes from established theory. Psychology's version rarely has a known correct curve to compare against. The skill here is different: living with genuinely noisy self-report and behavioural data, and resisting the causal story your own data will practically beg you to tell.

Why your own data teaches this better than a textbook figure

A published figure arrives finished, clean axes, a clear pattern, a caption that tells you what to see. It's a good way to learn to read a figure (article 04) and a poor way to learn what real psychological data looks like before someone smooths it into a publication.

Real self-tracked data is a mess in a specific, instructive way: single self-report items are noisy, day-to-day variation swamps small real effects, and the variables you're most interested in (mood, focus, stress) are exactly the ones with the least reliable measurement. Plotting your own numbers puts you in the position every psychology researcher is actually in, squinting at scatter, wondering whether that dip means anything, rather than the position of someone reading a result that's already been cleaned up.

There's also an advantage unique to using yourself as the subject: you know what actually happened on each day. A weird outlier isn't just a strange point; you know it was the day you had three coffees and a deadline. That context is exactly what a real researcher doesn't have about their participants and has to infer statistically instead.

The loop

1. Track something simple for two to three weeks. Sleep duration (from a phone or a written log), a daily mood rating on a fixed scale, or reaction time on a short online task, done at a consistent time of day.
2. Predict the relationship before plotting it. Do you expect more sleep to correlate with better next-day mood? Stronger, weaker, or no relationship than you'd guess?
3. Plot it yourself, choosing which variable goes on which axis and saying which you think would drive which, if either does.
4. Ask what the chart is hiding. A raw scatter of two three-week variables is almost always noisy enough that a real pattern and no pattern can look similar to the eye.
5. Ask what would be needed to call it more than a correlation. Almost always: more data, controls for other likely factors (what else varied across those weeks), and ideally a manipulation rather than passive tracking.

Here's my data: [paste, e.g. sleep hours and next-day mood rating, one row per day]. Plot it. Is there a visible relationship, and if so how strong does it look? What are at least two other factors that varied across these days and could produce this pattern without sleep driving mood directly? What would I need to add to this tracking to make a causal claim more defensible?

That middle question is article 06's method turned on data you generated yourself, which is where it's hardest to apply and most worth practising.

What two to three weeks of self-tracking can and can't tell you

It can show you: roughly how much your own numbers vary day to day (often more than you'd guess), whether an apparent pattern survives being plotted rather than just remembered, and what your own noise floor looks like, useful context for interpreting how "linked to" claims in published research compare to what a small personal dataset actually looks like.

It can't show you: anything causal, with this design. Sleep and mood both plausibly respond to a dozen other things that also varied across your three weeks, workload, exercise, alcohol, illness, the weather. A correlation in your own tracked data is exactly as vulnerable to a third variable as any correlational study you'd critique in coursework (article 06), and being your own data doesn't exempt it.

Across the variables worth tracking

Sleep and mood. The classic pairing, accessible, and genuinely variable enough to show something, while being confounded by everything that affects both.

Reaction time across the day. A short online reaction-time task done at three fixed points (morning, afternoon, evening) for a week or two often shows a real within-person pattern, and is closer to an experimental measure than a self-report one, which makes it a useful contrast case.

Screen time and next-day focus or mood, self-logged. Prone to the same third-variable problem as sleep and mood, and a good test case for the "what else varied" question.

Caffeine and a self-rated anxiety or focus scale. Fast-acting enough that same-day tracking, rather than next-day, is the more sensible design, a good prompt for the AI on study-design grounds before you even collect anything.

The thing to be honest about

Fourteen to twenty-one days of one person's data is enough to learn what noisy self-report data actually looks like, and not remotely enough to establish anything about yourself, let alone people in general. A write-up that says "my mood and sleep tracked together this month, though with n≈18 and several unmeasured factors I can't rule out a third variable" shows more understanding than one that says "this proves sleep controls my mood", and it's the same standard article 06 asks you to apply to published studies.

Pitfalls

1. Not predicting the relationship first. You'll look at the plot and feel you expected exactly that shape.
2. Skipping straight to a causal story. Your own data is the case where this temptation is strongest, because you have a theory about yourself already.
3. Treating a short personal dataset as more informative than it is. State the sample size and the obvious confounds in your own conclusion.
4. Tracking too many variables to log consistently. One or two, logged honestly every day, beats five logged sporadically.
5. The tell: your chart shows a clean pattern and you can't name two other things that varied across the same weeks.

Try this today

Start a two-variable log today, sleep and mood, or caffeine and focus, whichever you can log honestly for two weeks. Predict the relationship in writing before you start.

At the end of week two, plot it and ask what else varied across those days that could produce the pattern without either variable driving the other.