Research Methodology
Module-7
-Dr. Jay Prakash Verma
Ph.D., MBA, [Link]., [Link](H), UGC-NET
Associate Dean & Associate Professor- Author
AGENDA:
Parametric • Difference between parametric
and and non-parametric tests
Non-parametric Tests • Advantages and disadvantages
of non-parametric tests
tics • Tests: Chi-square test, ANOVA,
T-Test, Paired T-Test, Sign Test,
Wilcoxon Test, Mann Whitney U
Test, Kruskal-Wallis Test
Core Definitions
Parametric Tests Non-Parametric Tests
- Assume data follows a specific distribution (e.g., - No assumptions about data distribution
normal) - Analyze ranks, medians, or ordinal data
- Estimate population parameters (mean, variance) - Examples: Chi-square, Mann-Whitney U,
- Examples: T-Test, ANOVA, Paired T-Test Wilcoxon
• Advantages:
• Greater statistical power to
detect effects
• Handle unequal variances
Parametric between groups
• Support confidence intervals and
Tests regression modeling
• Disadvantages
• Require normality and
homogeneity of variance
• Sensitive to outliers
• Advantages
• - Robust to outliers and non-normal
data
Non- • - Analyze ordinal/ranked data (e.g.,
survey Likert scales)
Parametric • - No strict sample size requirements
Tests • Disadvantages
• - Less power to detect small effects
• - Require identical variability across
groups
• - Limited to hypothesis testing (no
confidence intervals)
Parametric Tests
• 1. Independent T-Test:
• - Compares means of two independent groups
• - Example: Testing if sales differ between two regions
• 2. Paired T-Test:
• - Compares means of related groups (e.g., pre/post-training)
• - Example: Employee productivity before vs. after a workshop
• 3. ANOVA:
• - Compares means across three+ groups
• - Example: Testing customer satisfaction across 5 store
branches
Non-Parametric Tests
1. Chi-Square Test:
- Tests independence between categorical variables
- Example: Analyzing if brand preference correlates with age groups
2. Mann-Whitney U Test:
- Non-parametric alternative to independent T-Test
- Example: Comparing median salaries of two departments with skewed data
3. Wilcoxon Signed-Rank Test:
- Non-parametric alternative to paired T-Test
- Example: Assessing median customer wait times before/after system upgrade
4. Kruskal-Wallis Test:
- Non-parametric alternative to ANOVA
- Example: Comparing median sales performance across 4 regions
5. Sign Test:
- Tests median differences in paired data with outliers
- Example: Evaluating if a policy change improved employee retention
When to Choose?
Decision Criteria
1. Use Parametric If:
• - Data is normally distributed (or sample size ≥30 for
CLT)
• - Research focuses on means and requires
confidence intervals
• - Groups have unequal variances
2. Use Non-Parametric If:
• - Data is ordinal, skewed, or has outliers
• - Sample size is small and normality uncertain
• - Median/rank is more meaningful than mean
• Scenario: Testing if a new ad campaign affects
Practical sales (data skewed).
- Parametric: Independent T-Test (if log-
Example transformed data normalizes)
- Non-Parametric: Mann-Whitney U Test
(retains original skewed data)
1. Chi-Square in Market Research:
- A/B testing website layouts for conversion rate
differences
Business 2. ANOVA in Product Development:
Applications: - Comparing average satisfaction scores for 3 product
variants
Case Studies 3. Wilcoxon in HR Analytics:
- Measuring median employee engagement scores
before/after remote work policy
Implementation
Tips