P-value > α → Fail to Reject H₀
Statistical Significance: Evidence that observed result is unlikely due to chance alone
Practical Significance: Whether observed effect is meaningful in real-world context
Effect Size: Magnitude of difference or relationship being tested
Confidence Interval: Range of plausible values for parameter
Statistical Power: Probability of detecting true effect
1. Define Research Question
2. State Hypotheses
3. Choose Test and Check Conditions
4. Collect Data and Calculate Statistics
5. Make Statistical Decision
6. Interpret Results in Context
Given: μ₀ = 200, n = 50, x̄ = 192, s = 15, α = 0.05
One-sample t-test since σ unknown and n > 30
H₀: μ ≥ 200 (drug doesn't lower cholesterol)
H₁: μ < 200 (drug lowers cholesterol)
t = (x̄ - μ₀)/(s/√n) = (192 - 200)/(15/√50)
t = -8/(15/7.07) = -8/2.12 = -3.77
df = 49, left-tailed test
p-value = P(T₄₉ ≤ -3.77) ≈ 0.0002
Since p-value (0.0002) < α (0.05), reject H₀
There is strong evidence the drug lowers cholesterol
Mean reduction: 8 mg/dL (practically significant?)
Reject H₀. The drug appears effective at lowering cholesterol (p < 0.001).
• One-sample t-test: When σ unknown
• Medical Context: Left-tailed for beneficial effect
• Decision Rule: p-value vs α comparison
Given: μ₀ = 10.0, n = 36, x̄ = 10.1, s = 0.3, α = 0.01
H₀: μ = 10.0 (process is on target)
H₁: μ ≠ 10.0 (process is off target)
t = (10.1 - 10.0)/(0.3/√36) = 0.1/(0.3/6) = 0.1/0.05 = 2.0
df = 35, two-tailed test
p-value = 2 × P(T₃₅ ≥ 2.0) ≈ 2 × 0.027 = 0.054
Since p-value (0.054) > α (0.01), fail to reject H₀
No significant evidence to adjust the process
Observed difference could be due to random variation
Fail to reject H₀. No significant evidence to adjust the process (p = 0.054).
• Two-tailed Test: Detecting any deviation from target
• Quality Control: Conservative α level for process adjustment
• Cost Consideration: Avoid unnecessary adjustments
Given: μ₀ = 75, σ = 10 (known), n = 25, x̄ = 78, α = 0.05
H₀: μ ≤ 75 (method doesn't improve scores)
H₁: μ > 75 (method improves scores)
Since σ known, use z-test: z = (x̄ - μ₀)/(σ/√n)
z = (78 - 75)/(10/√25) = 3/(10/5) = 3/2 = 1.5
Right-tailed test: p-value = P(Z ≥ 1.5) = 0.0668
Since p-value (0.0668) > α (0.05), fail to reject H₀
No significant evidence the new method improves scores
However, 3-point improvement might be practically meaningful
Fail to reject H₀. No significant evidence of improvement (p = 0.067).
• Z-test: When σ is known
• Right-tailed: Testing for improvement
• Practical vs Statistical: Consider both significance types
Given: p₀ = 0.60, n = 200, x = 130, p̂ = 130/200 = 0.65, α = 0.01
H₀: p = 0.60 (claim is accurate)
H₁: p ≠ 0.60 (claim is inaccurate)
np₀ = 200 × 0.60 = 120 ≥ 5 ✓
n(1-p₀) = 200 × 0.40 = 80 ≥ 5 ✓
z = (p̂ - p₀)/√[p₀(1-p₀)/n]
z = (0.65 - 0.60)/√[0.60 × 0.40/200]
z = 0.05/√[0.24/200] = 0.05/√0.0012 = 0.05/0.0346 = 1.44
Two-tailed test: p-value = 2 × P(Z ≥ 1.44) = 2 × 0.0749 = 0.1498
Since p-value (0.1498) > α (0.01), fail to reject H₀
No significant evidence against the company's claim
Survey results consistent with 60% preference
Fail to reject H₀. Survey results consistent with 60% claim (p = 0.150).
• Proportion Test: For categorical data
• Success-Failure Condition: Check np₀ ≥ 5 and n(1-p₀) ≥ 5
• Marketing Context: Conservative testing for claims
Given: p₀ = 0.30, n = 500, x = 160, p̂ = 160/500 = 0.32, α = 0.05
H₀: p ≥ 0.30 (smoking rate ≥ 30%)
H₁: p < 0.30 (smoking rate < 30%)
np₀ = 500 × 0.30 = 150 ≥ 5 ✓
n(1-p₀) = 500 × 0.70 = 350 ≥ 5 ✓
z = (0.32 - 0.30)/√[0.30 × 0.70/500]
z = 0.02/√[0.21/500] = 0.02/√0.00042 = 0.02/0.0205 = 0.98
Left-tailed test: p-value = P(Z ≤ 0.98) = 0.8365
Since p-value (0.8365) > α (0.05), fail to reject H₀
No significant evidence that smoking rate is less than 30%
Sample suggests rate might be higher than 30%
Fail to reject H₀. No evidence smoking rate is less than 30% (p = 0.837).
• Public Health Context: Important policy implications
• Left-tailed Test: Testing for "less than" claim
• Large Sample: Normal approximation valid