Hypothesis testing is one of the most testable (and most misunderstood) topics in statistics. It shows up in exam questions, research papers, A/B tests, and everyday claims like “this new method works better.” This guide gives you a clean mental model for hypothesis tests, explains p-values and confidence intervals in plain language, and highlights the traps that cause the most mistakes.
Start with the real question, not the formula
A hypothesis test is a structured way to answer: “Is the observed difference (or relationship) likely to be real, or could it be explained by random variation?” The goal isn’t to “prove” something true—it’s to evaluate whether the data are surprising under a specific assumption.
Null vs. alternative: what you’re actually comparing
Every test begins with two competing statements:
- Null hypothesis (H₀): the default claim, usually “no effect,” “no difference,” or “no association.”
- Alternative hypothesis (H₁ or Hₐ): what you suspect or want evidence for—an effect, difference, or association.
Example: If you’re comparing two teaching methods, a common setup is:
H₀: average scores are equal
Hₐ: average scores differ
Test statistic: compressing the evidence into one number
Most hypothesis tests compute a test statistic that summarizes how far your sample result is from what H₀ predicts, scaled by expected variability. Different tests use different statistics (z, t, χ², F), but the core idea is consistent: the further from H₀, the stronger the evidence against it.
What a p-value really means (and what it doesn’t)
A p-value is the probability of observing results at least as extreme as yours if the null hypothesis were true.
- Small p-value: strong evidence against H₀
- Large p-value: insufficient evidence to reject H₀
Common misconceptions:
- ❌ p is the probability H₀ is true
- ❌ large p-value proves no effect
- ❌ p < 0.05 means the result is important
✔ Reality: p-values measure surprise under H₀, not truth or importance.
Significance level (α): your false-alarm threshold
Before analyzing data, you choose a significance level (α), often 0.05.
- If p ≤ α → reject H₀
- If p > α → fail to reject H₀
α represents how often you’re willing to make a false positive.
Type I and Type II errors
Hypothesis testing involves two possible mistakes:
- Type I error: rejecting H₀ when it’s true (false positive)
- Type II error: failing to reject H₀ when it’s false (false negative)
The probability of detecting a true effect is called power (1 − β).
Confidence intervals: more than “significant or not”
A confidence interval (CI) gives a range of plausible values for a parameter.
CI=θ^±z∗⋅SE
Key insight:
- If a 95% CI does not include 0, it aligns with rejecting H₀ at α = 0.05 (in many cases)
- CIs show effect size + uncertainty, not just a yes/no decision
One-tailed vs. two-tailed tests
- Two-tailed: checks for any difference (default choice)
- One-tailed: checks for a specific direction
Rule: never choose the tail direction after seeing the data.
Assumptions matter
Using the correct test requires checking:
- Independence (most critical)
- Normality (small samples)
- Equal variances (in some comparisons)
- Correct data type (means vs proportions vs ranks)
If assumptions fail, use alternatives (e.g., Welch’s test, nonparametric methods).
Practical significance: what p-values miss
Statistical significance ≠ real-world importance.
Always consider:
- Effect size
- Confidence interval width
- Context (does it matter?)
A tiny effect can be “significant” with large data, while a meaningful effect may not be detected in small samples.
A quick exam-ready checklist
- State H₀ and Hₐ (with direction)
- Identify the parameter
- Check assumptions
- Interpret test statistic and p-value
- Compare p with α
- Answer in context (no “prove”)
- Include CI and effect size
Keep learning
- https://cursa.app/free-online-courses/statistics-basics
- https://cursa.app/free-online-courses/applied-statistics
- https://cursa.app/free-online-courses/statistics-for-data-science
- https://cursa.app/free-courses-basic-studies-online
- https://cursa.app/free-online-basic-studies-courses
- https://www.khanacademy.org/math/statistics-probability
