ANOVA & Chi-Square: Hypothesis Testing for Group Differences & Categorical Data

Core Concepts: What ANOVA and Chi-Square Solve

1. Analysis of Variance (ANOVA): Testing Group Differences

ANOVA is used to determine whether three or more population means are equal. It compares the variability between groups to the variability within groups.

Key Idea: Partitioning Variance

ANOVA splits total variability into:

• Between-group variance (explained by group differences)
• Within-group variance (random error within groups)

If the between-group variance is significantly larger than the within-group variance, we reject the null hypothesis that all group means are equal.

Assumptions of ANOVA

mindmap
  root((ANOVA Assumptions))
    Normality["Data in each group is normally distributed"]
    Homogeneity["Variances (σ²) are equal across groups (Homogeneity of Variance)"]
    Independence["Observations are independent (no repeated measures)"]
    Randomness["Samples are randomly selected"]

Types of ANOVA

2. Chi-Square Test: Analyzing Categorical Data

Chi-Square tests whether observed frequencies differ from expected frequencies in categorical data.

Key Idea: Goodness-of-Fit vs. Test of Independence

Assumptions of Chi-Square

mindmap
  root((Chi-Square Assumptions))
    Independence["Observations are independent"]
    Expected["Expected frequency ≥5 in at least 80% of cells (for 2×2 tables)"]
    Categorical["Data is categorical (nominal/ordinal)"]

Step-by-Step Worked Examples

Example 1: One-Way ANOVA (Testing Fertilizer Effect on Crop Yield)

Scenario: A farmer tests three fertilizers (A, B, C) on 15 plots (5 per fertilizer). Yields (kg/plot) are:

Step 1: Calculate Group Means and Grand Mean

# Means per group
A_mean = (45+48+42+50+47)/5 = 46.4
B_mean = (50+52+48+55+53)/5 = 51.6
C_mean = (40+42+38+45+41)/5 = 41.2

# Grand mean
Grand_mean = (46.4*5 + 51.6*5 + 41.2*5)/15 = 46.4

Step 2: Calculate Sum of Squares (SS) Calculations:

• SSB = 5[(46.4-46.4)² + (51.6-46.4)² + (41.2-46.4)²] = 108.8
• SSW = Σ[(45-46.4)² + ... + (41-41.2)²] = 112.8
• SST = SSB + SSW = 221.6

Step 3: Compute Mean Squares (MS) and F-statistic

Step 4: Compare F to Critical Value

• Critical F (α=0.05, df1=2, df2=12) ≈ 3.89 (from F-table).
• Since 5.79 > 3.89, reject H₀. Fertilizers have a significant effect.

A labelled ANOVA table showing SSB, SSW, MS, and F-statistic. (Image: Luyan Li, CC BY 4.0, via Wikimedia Commons)

Example 2: Chi-Square Test of Independence (Education vs. Voting Preference)

Scenario: A survey of 200 voters shows:

Step 1: State Hypotheses

• H₀: Education level and voting preference are independent.
• H₁: They are dependent.

Step 2: Calculate Expected Frequencies For High School & Party X: Expected = (Row Total × Column Total) / Grand Total = (80 × 100) / 180 ≈ 44.44

Step 3: Compute Chi-Square Statistic

Step 4: Compare to Critical Value

• Critical Chi² (α=0.05, df=(2-1)(3-1)=2) ≈ 5.99.
• Since 14.11 > 5.99, reject H₀. Education level affects voting preference.

In the Real World

1. E-Sewa & Khalti: Testing User Experience Across Payment Methods

• ANOVA Application: E-Sewa tests whether three payment methods (credit card, mobile wallet, bank transfer) affect transaction success rates.
  • Groups: Payment methods (independent variable).
  • Dependent Variable: Time taken to complete payment (in seconds).
  • Business Use: If one method is significantly faster, E-Sewa can promote it to reduce dropout rates.

2. Daraz: Chi-Square Analysis of Customer Reviews by Product Category

• Chi-Square Application: Daraz analyzes whether product category (electronics, fashion, groceries) is independent of review sentiment (positive/negative).
  • Example: If electronics have more negative reviews than expected, Daraz can investigate quality control or adjust return policies.

3. NTC & Ncell: Comparing Network Performance Across Regions

• Two-Way ANOVA Application: NTC tests if network speed varies by:
  • Region (Kathmandu, Pokhara, Chitwan) (Factor 1).
  • Time of Day (Morning, Evening, Night) (Factor 2).
  • Business Use: If Pokhara has slower speeds at night, NTC can invest in infrastructure there.

4. NEPSE: Testing Stock Price Movements by Sector

• ANOVA Application: NEPSE compares monthly returns of stocks in:
  • Banking sector.
  • Hydro sector.
  • IT sector.
  • Business Use: If one sector outperforms others significantly, investors can shift portfolios accordingly.

Common Mistakes & How to Avoid Them

Exam Tip

What Examiners Look For

1. Correct Hypothesis Formulation

   • Always write H₀ and H₁ clearly (e.g., "All group means are equal" vs. "At least one differs").
   • For Chi-Square: "Variables are independent" vs. "Variables are dependent."

2. Step-by-Step Calculations

   • Show SSB, SSW, SST for ANOVA.
   • For Chi-Square, list observed vs. expected and compute (O-E)²/E for each cell.

3. Decision Rule

   • Compare calculated F/Chi² to critical value (or use p-value).
   • State whether to reject H₀ and interpret in context.

4. Assumption Checks

   • Mention if data meets normality, homogeneity of variance (Levene’s test), or expected frequency ≥5.

5. Post-Hoc Analysis (ANOVA Only)

   • If ANOVA is significant, mention Tukey’s test (even if not calculated).

Quick Revision Checklist

• Can I partition variance into between/within groups?
• Do I know when to use ANOVA vs. t-test (ANOVA for ≥3 groups)?
• Can I compute expected frequencies for Chi-Square?
• Do I interpret p-values correctly (p < 0.05 → reject H₀)?

Final Note: ANOVA and Chi-Square are powerful tools for business decisions—from marketing strategies (Khalti promotions) to operational efficiency (NTC network upgrades). Master the calculations and assumptions, and you’ll ace the exam!