Blog / Why the Most Extreme Self-Test Dimension Is Not a Significant Finding

Why the Most Extreme Self-Test Dimension Is Not a Significant Finding

The highest dimension can guide reflection, but it is not a hypothesis-test result.

October 15, 2026 · 4 min read · By Cepingku Editorial Team

A multidimensional report always has a highest and lowest result. That can prompt useful reflection, but it does not establish a statistical difference or predict that the same dimension will remain most extreme. Statistical significance requires a pre-specified hypothesis, sampling model, test, and threshold. This site does not claim to run significance tests on its reports; this article addresses a mistake readers can make when interpreting research or many indicators.

A fictional ten-test teaching model

Assume ten independent tests, every null hypothesis is true, and each test has alpha = .05. The probability of no false positive in one test is .95. The probability of no false positive in all ten is .95^10. By the complement rule, the probability of at least one false positive is 1−.95^10 ≈ .401, or 40.1 percent. This is direct probability arithmetic for a deliberately fictional independent model; it is not evidence that this site runs ten tests or has a 40.1 percent error rate.

Real dimensions are often correlated, and real tests may not be independent, so 40.1 percent cannot be copied onto an actual report. The lesson is narrower: after searching many results and selecting the most interesting one, a single .05 interpretation no longer describes the error risk of the entire search.

Plan the family of comparisons

A stronger research workflow states the main hypotheses and comparison family before inspecting results, then reports all tests rather than only the smallest p value. Benjamini and Hochberg’s original 1995 paper introduced false discovery rate (FDR), the expected proportion of falsely rejected hypotheses, and proved a sequential procedure under independent test statistics. That source supports the multiple-testing problem and an FDR method; it is not the source of the complement calculation above. Exploratory work is still useful when it is labelled as exploratory: a standout result can motivate a new, pre-specified study or a transparent follow-up analysis. It should not silently become confirmation merely because it was selected after many alternatives were viewed. The correction method, the intended error rate, the family of comparisons, and any dependence assumptions should be stated with the result so another reader can see what claim is being controlled.

For self-reflection, use language such as ‘relatively high in this set of responses’ and check it against situations, item content, and later behaviour. The Big Five Personality Test organises one set of self-reported responses across five dimensions. It is not presented here as a multiple-testing procedure and this article does not claim it supplies significance labels, FDR, norms, or diagnostic conclusions.

Want to see where you land?

Take the Big Five Personality Test and get your results in minutes.

Take the Big Five Personality Test →

← Back to the blog