#6a Advance Hypothesis Testing

🎓 Advanced Hypothesis Testing

University-Level Statistical Analysis for Civil Engineering
Comprehensive 120-Minute Lecture | Theoretical Foundations + Advanced Applications

📢 Lecture Introduction & Learning Objectives

Welcome to Advanced Hypothesis Testing! This lecture equips you with the statistical decision-making framework used in structural design, geotechnical analysis, and quality control.

Core Question: "Given limited sample data, can we confidently reject design specifications or accept engineering assumptions?"

🎯 University Learning Outcomes

• Formulate and solve hypothesis tests for means (z, t) and variances (χ², F)
• Derive test statistics from sampling distributions
• Apply power analysis and sample size determination
• Interpret Type I/II errors in civil engineering risk contexts
• Conduct non-parametric alternatives when normality fails

🎯 Module 1: Theoretical Foundations (20 mins)

1.1 Neyman-Pearson Lemma & Decision Theory

The optimal test maximizes power (1-β) for fixed α using likelihood ratio:

Λ = sup_θ∈Θ₀ L(θ|x) / sup_θ∈Θ L(θ|x) ≶ k

1.2 Sampling Distributions & Pivotal Quantities

Central Limit Theorem: √n(x̄ - μ)/(σ/√n) → N(0,1)
t-Distribution: (x̄ - μ)/(s/√n) → t_n-1
χ²-Distribution: Σ[(X_i-μ)²/σ²] → χ²_n-1

1.3 Comprehensive Error Framework

	H₀: μ = μ₀ True		H₀: μ ≠ μ₀ True
	Prob	Decision	Prob	Decision
Region A	1-α	Accept	β	Accept
Region B	α	Reject	1-β	Reject

📊 Module 2: Advanced Tests for Means (30 mins)

2.1 Z-Test: Large Sample Theory & Applications

Z = √n(x̄ - μ₀)/σ ~ N(0,1)
Rejection Region: |Z| > z_α/2 (two-tailed)
Power: 1-β = P(Z > z_α - √n|μ-μ₀|/σ)

Bridge Load Analysis (n=50):
x̄=422kN, σ=58kN, μ₀=400kN, α=0.05
Z = √50(422-400)/58 = 2.78 > 1.96
p-value: P(Z>2.78) = 0.0027 < 0.05
Power Analysis: For μ=410kN, power=0.87

2.2 T-Test: Small Sample Exact Distribution

T = √n(x̄ - μ₀)/s ~ t_n-1
Exact Distribution: Derived from Normal/χ² independence

Concrete Compressive Strength (n=12):
Data: [27.8, 29.2, 30.1, 28.9, 29.7, 30.4, 28.5, 29.8, 30.2, 29.1, 28.7, 29.9]
x̄=29.42MPa, s=0.78MPa, μ₀=30MPa, α=0.05
t₁₁ = √12(29.42-30)/0.78 = -2.45
t_crit(0.025,11) = ±2.201 → REJECT H₀
Conclusion: "Batch fails M30 specification"

2.3 Paired T-Test & Two-Sample Analysis

Paired: t = (d̄ - μ_D)/(s_D/√n)
Two-Sample: t = (x̄₁-x̄₂)/√(s_p²(1/n₁+1/n₂))

📈 Module 3: Variance Tests - Theory & Applications (25 mins)

3.1 Chi-Square Test: Exact Distribution

χ² = (n-1)s²/σ₀² ~ χ²_n-1
Two-tailed: χ²_1-α/2 < χ² < χ²_α/2
One-tailed (σ²>σ₀²): χ² > χ²_1-α

Geotechnical Soil Test (n=15):
Shear strength s²=3.84kPa², σ₀²=2.25kPa² (CV=25%)
χ²₁₄ = 14×3.84/2.25 = 23.87
Critical: χ²_0.025,14=5.63, χ²_0.975,14=26.12
5.63 < 23.87 < 26.12 → FAIL TO REJECT
Design Decision: "Soil variability acceptable for foundation"

3.2 F-Test: Ratio of Independent χ² Variables

F = (s₁²/σ₁²)/(s₂²/σ₂²) = s₁²/s₂² ~ F_{n₁-1,n₂-1}
Decision: F > F_{α,n₁-1,n₂-1}

🏗️ Module 4: Civil Engineering Applications & Standards (20 mins)

Test Type	Civil Application	IS Code Reference	α Level	Decision Rule
Z-Test	Traffic/Environmental Loads	IRC:6-2017	0.05	Load > Design
T-Test	Concrete Quality Control	IS:456-2000	0.05	Strength < Spec
χ²-Test	Soil Property Variability	IS:2720	0.10	CV > Limit
F-Test	Batch/Material Comparison	IS:10262	0.05	Variances unequal

Quality Control Protocol (IS:456):
M30 Concrete: Test 3 cubes at 7d, 6 at 28d
H₀: μ ≥ 30MPa vs H₁: μ < 30MPa
Reject if mean of any group < characteristic strength

🎯 Module 5: Power Analysis & Sample Size (15 mins)

📐 Sample Size Formulas

Z-Test: n = [z_α/2 + z_β]²(σ/δ)²
T-Test: n ≈ [t_α/2,df + t_β,df]²(σ/δ)²
Concrete Example: μ₀=30MPa, δ=1MPa, σ=2MPa, Power=0.9
n = [1.96+1.28]²(2/1)² = 39.2 → 40 samples minimum

✍️ Advanced In-Class Exercises

Exercise 1 - Structural Steel Yield Strength:
n=20, x̄=245MPa, s=8MPa, Spec=250MPa, α=0.05
Solution: t₁₉=-2.50 > -2.093 → Acceptable

Exercise 2 - Aggregate Variance Test:
n=25, s²=16.4, σ₀²=9 (CV=30%), α=0.10
Solution: χ²₂₄=43.11 ∈ [13.85, 36.42] → Acceptable

🧠 University-Level Assessment (10 Questions)

1. The Neyman-Pearson lemma optimizes: a) Power for fixed α
2. T-distribution approaches normal when: a) df → ∞
3. χ²_n-1 expected value equals: a) n-1
4. Power = 1-β measures: a) Correct rejection probability
5. For concrete QC, typical α = a) 0.05

🚀 Master Equation Summary

General Test Statistic: T = [Sample Stat - Hypothesized Value] / SE
Decision Rule: |T| > Critical Value or p-value < α
Engineering Action: Translate statistical decision → Design/Quality decision

Advanced Homework: Design a sampling plan for M40 concrete validation (n=?, Power=0.95)

Research Project: Compare Type I/II error costs in foundation design failure scenarios

Next: Non-parametric Tests + Bayesian Decision Theory | B.Tech Civil Engineering | March 2026

👨‍🏫 Developed for University-Level Civil Engineering Curriculum