Correlation Is Not Causation: How to Read Your Mood Data Correctly

The moment the app "found something"

You have been tracking your mood for a few weeks. Then the analysis shows you a connection: "On days with coffee, your mood is higher." The first reflex is obvious. More coffee, better mood.

Unfortunately, that is not how it works. The sentence describes a correlation, a shared occurrence. About the cause it says nothing. This is exactly where the most common false conclusions in self-tracking happen, and exactly where it is decided whether your data helps you or leads you astray.

This article is the entry-level data literacy for anyone who measures mood, sleep or factors. Not a statistics lecture, but four reasoning errors and how to avoid them.

What correlation means, and what it does not

A correlation measures how strongly two quantities change together. The value lies between minus one and plus one. There is nothing more to it. A correlation contains no direction and no cause, it is a pure observation of the form "these two things move together."

The best-known evidence for this is Tyler Vigen's Spurious Correlations. There, margarine consumption correlates almost perfectly with the divorce rate in Maine. No one would claim that one causes the other. The lines run parallel by chance.

In science, too, this distinction is a topic of its own. The statisticians Naomi Altman and Martin Krzywinski put it in a formula in Nature Methods: correlation means association, but not causation. In your data it is just less obvious.

The four most common false conclusions

1. Reversed direction. The app says: "On days with a lot of exercise, your mood is better." The opposite direction is just as plausible. On days when you already feel good, you are more likely to go out and exercise. Then the good mood produces the exercise, not the other way around. Usually both directions act at once.

2. The hidden third factor. The trickiest error. Two things correlate because a third, unmeasured one drives both. Your mood is higher on coffee days? The true cause may be sleep. After a good night you are more awake, have more drive and enjoy your coffee. Sleep is the confounder, the coffee just comes along for the ride.

3. Chance with little data. With ten days, some connection almost always shows up, simply by chance. The more factors you analyze at once and the fewer days you have, the higher the chance that a pattern is pure noise. How many days you need is clarified by (article: innerpulse/blog/2026/07/how-many-days-mood-tracking-patterns text: How many days of tracking until patterns appear).

4. The story after the fact. Our brain finds a story for almost any data point. "Of course, that is why Tuesday was so bad." These stories convince because they arise after the data and fit perfectly. The test is: would you also have predicted it beforehand?

How to draw solid conclusions anyway

Correlation is not worthless, it is the beginning. Here is how to make more of it:

The strongest lever for a private person is the experiment with yourself. When you deliberately change a factor instead of just observing it, you get closer to a real cause-and-effect test. Action produces data, mere observation does not. And pay attention to time: when poor sleep on Monday is connected to low mood on Tuesday, the direction is at least plausible. More on this in (article: innerpulse/blog/2026/01/how-sleep-affects-your-mood text: How sleep affects your mood).

Why InnerPulse speaks of "observation"

This is exactly why InnerPulse phrases its analyses as observations, not as diagnoses. A sentence like "On days with exercise you ate well 40 percent more often" is deliberately descriptive. It claims no cause, it invites you to think further yourself. How the app forms these patterns is explained by (article: innerpulse/blog/2026/02/recognizing-mood-patterns text: Recognizing patterns in your mood), and the bigger picture is provided by the (article: innerpulse/blog/2026/04/innerpulse-guide text: InnerPulse guide).

This caution is not a lack of confidence, but the honest treatment of what observational data can deliver. Anyone who promises more is selling you a certainty that the data does not provide.

The one sentence to take away

The next time your app shows you a connection, mentally replace "causes" with "is connected to" and add a question: "What would happen if I changed it?" This small rephrasing turns a seductive half-truth into a tool. Your data rarely proves anything, but it reliably shows you where it is worth taking a closer look.

This article does not replace medical advice. It helps you better classify your own observations.

Further reading

• (article: innerpulse/blog/2026/07/how-many-days-mood-tracking-patterns text: How many days of tracking until patterns appear) clarifies the question of enough data.
• (article: innerpulse/blog/2026/02/recognizing-mood-patterns text: Recognizing patterns in your mood) shows how to interpret connections in practice.
• (article: innerpulse/blog/2026/04/90-day-mood-tracking-field-report text: 90 days of mood tracking: a field report) shows how patterns emerge over time.
• (article: innerpulse/blog/2026/03/why-streaks-harm-people-with-depression text: Why streaks harm people with depression) explains why more data is not automatically better.
• Scientific background: Altman & Krzywinski, Nature Methods (2015)
• Illustrative example: Spurious Correlations by Tyler Vigen