Correlation in a Personal Log and Why It Misleads
A personal log can show two measures moving together for months and still be describing nothing causal. That is the central caution in the literature on self-tracking: correlation, especially in short records with many variables, is a description of coincidence unless a plausible mechanism and a controlled comparison support it.
Why small records produce confident-looking patterns
Someone tracking step count, sleep quality and fasting glucose on a home glucose meter generates dozens of paired values within weeks, and any two of them can drift in the same direction by chance. Statistical guidance from research-methods sources notes that with roughly twenty to thirty overlapping observations, a moderately strong correlation can appear from noise alone, particularly when the series are both trending — a gradual rise in weekly steps alongside a seasonal change in vitamin D level, for instance. Sleep apnoea screening, screen curfew timing and stress load are frequently logged in the same period, which multiplies the number of pairs that can be compared. The more columns a log carries, the more likely one pairing looks meaningful. What no correlation coefficient can supply is direction: waking up unrested may reduce activity, or reduced activity may worsen sleep, and the log records both without distinguishing them.
What a log can and cannot establish
These pages treat self-tracking as a way to generate questions rather than answers. A pattern worth noting — say, lower energy levels on days following a late meal — becomes useful only when it is tested against a deliberate change held steady long enough to see whether the relationship survives. That is where artefacts such as a myth-vs-evidence table help: they force a distinction between an association observed once and a relationship reproduced across separate periods. Adherence rate matters here too, because a log kept only on good days is a biased sample. The honest reading of most personal correlations is that they narrow what to look at, not what is true.
The practical value of a personal log lies in the questions it raises for a clinician or a lab panel to examine, not in the conclusions it appears to offer on its own.
