Hypothesis
Name variants
- English
- Hypothesis
- Kanji
- 仮説
Quality / Updated / COI
- Quality
- Reviewed
- Updated
- Source
- Citations & Trust
- COI
- none
TL;DR
A hypothesis is a testable statement that can be evaluated with data, often framed as null and alternative hypotheses.
Definition
A hypothesis defines a specific claim about a population or relationship that can be tested through evidence. In statistical testing, the null hypothesis represents no effect, while the alternative represents a meaningful difference or relationship. Clear hypotheses guide experiment design, sample size decisions, and interpretation of results.
Decision impact
- It determines the experiment design and what data are required.
- It shapes which metrics and thresholds indicate success or failure.
- It influences how confidently results can be acted upon.
Key takeaways
- State hypotheses in measurable terms with defined variables.
- Specify null and alternative hypotheses before testing.
- Choose sample sizes that can detect meaningful effects.
- Interpret results in context, not just by p-values.
- Document assumptions so others can replicate the test.
Misconceptions
- A hypothesis is not a casual guess; it is a testable statement.
- Failing to reject the null does not prove the null is true.
- Changing hypotheses after seeing data undermines validity.
Worked example
A product team tests whether a new onboarding flow increases activation. The null hypothesis states there is no difference, and the alternative states activation increases by at least 5%. They run an A/B test with a sample size large enough to detect the effect. Results show a statistically significant increase, and the team rolls out the change while documenting assumptions and limitations.
Citations & Trust
- Introductory Statistics 2e 9.1 Null and Alternative Hypotheses (OpenStax)