Null Hypothesis Significance Testing – Quick Review
Hypothesis Testing Framework
- Start with two competing statements about a population parameter (usually \mu)
- Assume the null hypothesis is true; use sample data to challenge it
- Goal: decide whether to reject H0 (supporting H1) or fail to reject H_0
Null vs. Alternative Hypotheses
- Always created as a pair
- H_0 (null/comparison): contains an equality (e.g., =, \le, \ge)
- H_1 (alternative/research): contains the opposite inequality (e.g., \neq,
- Must be mutually exclusive and collectively exhaustive
- Example (non-directional):
- H0: \mu{NYC} = 2.17
- H1: \mu{NYC} \neq 2.17
Directionality
- Non-Directional (Two-Tailed): test for any difference; H_1 uses \neq
- Directional (One-Tailed): test for a specific direction; H_1 uses > or <
- Example (greater-than): H0: \mu{Law} \le 21, H1: \mu{Law} > 21
Sampling Distribution & Standard Error
- Under H_0, the sampling distribution of \bar x is normal
- Mean = \mu_{0}
- Standard error = \sigma/\sqrt{N}
- Provides the reference to judge how extreme the sample mean is
Alpha (\alpha), p-Value, and Critical Values
- Alpha (\alpha): researcher-set rejection threshold (commonly 0.05)
- Critical value (CV): cutoff on the H_0 distribution that bounds the rejection region
- One-tailed: single CV in the tail with area \alpha
- Two-tailed: two symmetric CVs with total area \alpha (each tail \alpha/2)
- p-value: probability of obtaining the observed statistic (or more extreme) given H_0 is true
- Reject H_0 if p < \alpha (statistically significant)
Decision Outcomes
- Reject H0: evidence favors H1; result is statistically significant
- Fail to Reject H0: insufficient evidence against H0; result is not significant
Step-by-Step Summary
- State H0 and H1 (include directionality)
- Select \alpha
- Determine sampling distribution & critical value(s)
- Collect sample, compute statistic (e.g., \bar x)
- Compare statistic to CVs or compute p-value
- Make decision: reject or fail to reject H_0
Key Vocabulary
- Null Hypothesis (H_0): status-quo/comparison claim
- Alternative Hypothesis (H_1): researcher’s substantive claim
- Alpha (\alpha): Type I error rate (probability of wrongly rejecting H_0)
- p-value: observed extremity probability under H_0
- Critical Value: boundary of the rejection region
- Statistical Significance: when p < \alpha (or statistic falls in rejection region)