TABLE OF CONTENTS

Chi-Square Test 3
What is Chi-Square? 3
Chi-Square Formula 4
Types of Chi-Square Tests 4
Step-by-Step Example 5
Hands-On Activity 7
ANOVA (Analysis of Variance) 8
What is ANOVA? 8
One-Way ANOVA 9
ANOVA Formula & Table 10
Step-by-Step Example 11
Post-Hoc Tests 13
Hands-On Activity 13
P-Value 14
What is a P-Value? 14
Interpreting P-Values 15
Common Misconceptions 16
Hands-On Activity 17
Tables & Critical Values 18
What are Critical Values? 18
How to Read Statistical Tables 19
Chi-Square Table 19
F-Distribution Table (ANOVA) 21
Decision Framework 22
Hands-On Activity 22

Statistics for Educators • Page 1 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 SECTION
CHI-SQUARE TEST (χ²)
Testing Relationships Between Categorical Variables

What is the Chi-Square Test?

The Chi-Square test is a statistical test used when your data is in categories (like Yes/No, Colors,
Types of food, etc.). It answers one fundamental question:

💡 The Core Question Chi-Square Answers

"Is there a real relationship between two categorical variables, OR did the pattern I see
happen just by chance?"

Imagine you toss a coin 100 times and get 60 Heads and 40 Tails. Is your coin
ANALOGY
unfair, or was it just luck? Chi-Square helps you decide — statistically!

Real-Life Scenario
Scenario: A teacher wants to know: "Does gender affect a student's choice of subject (Science vs.
Arts)?"

She surveys 200 students and records their gender and chosen subject. The results look like this:

Gender / Subject Science Arts Row Total

Male 70 30 100
Female 40 60 100
Column Total 110 90 200

Looking at the table, males seem to prefer Science and females seem to prefer Arts. But is this
difference real, or could it be due to random chance? Chi-Square will tell us.

Statistics for Educators • Page 2 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

The Chi-Square Formula

χ² = Σ [ (O - E)² / E ]
Where: O = Observed frequency | E = Expected frequency | Σ = Sum of all cells

Breaking it down:
• O (Observed): The actual count you collected in your survey/experiment
• E (Expected): The count you would EXPECT if there was NO relationship between variables
• (O - E): The difference between what you got and what you expected
• (O - E)²: Squaring removes negative signs — all differences become positive
• ÷ E: Dividing by E makes it proportional (a difference of 10 in a group of 20 is very different
from 10 in a group of 1000)
• Σ: Add up all these values from every cell in your table

Calculating Expected Values

E = (Row Total × Column Total) / Grand Total

This is the 'expected' count IF there were no relationship

Using our example:

• E (Male, Science) = (100 × 110) / 200 = 55
• E (Male, Arts) = (100 × 90) / 200 = 45
• E (Female, Science) = (100 × 110) / 200 = 55
• E (Female, Arts) = (100 × 90) / 200 = 45

Step-by-Step Chi-Square Calculation

Cell O (Observed) E (Expected) (O-E)² (O-E)²/E

Male/Science 70 55 (70-55)²=225 225/55 = 4.09

Male/Arts 30 45 (30-45)²=225 225/45 = 5.00

Female/Science 40 55 (40-55)²=225 225/55 = 4.09

Female/Arts 60 45 (60-45)²=225 225/45 = 5.00

χ² = 18.18 ✓

χ² = 4.09 + 5.00 + 4.09 + 5.00 = 18.18

Statistics for Educators • Page 3 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

Degrees of Freedom (df)

df = (number of rows - 1) × (number of columns - 1)

For our 2×2 table: df = (2-1) × (2-1) = 1

Degrees of freedom represents how many values can vary freely once we know the totals. Think of
it as: in a table with fixed margins, how many cells can you freely fill in?

Making the Decision

With df = 1 and α = 0.05 (the most common significance level), the critical value from the Chi-
Square table = 3.841

If χ² calculated > χ² critical → REJECT the null hypothesis (H₀)

If χ² calculated < χ² critical → FAIL TO REJECT H₀
DECISION
RULE
Our result: 18.18 > 3.841 → We REJECT H₀
Conclusion: There IS a significant relationship between gender and subject
choice.

Types of Chi-Square Tests

Test Type When to Use Real-Life Example

Goodness of Fit Testing if observed data fits an Does a dice roll have equal
expected distribution probability for all 6 faces?
Test of Independence Testing if 2 categorical variables are Is vaccine preference related to age
related group?
Test of Homogeneity Comparing distributions across Do different cities have the same
different populations distribution of political views?

Statistics for Educators • Page 4 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

Assumptions of Chi-Square

⚠️ Important Assumptions — Violating These Invalidates Your Test

• Data must be in COUNTS (frequencies), not percentages or means
• Each observation must be INDEPENDENT (one person/item counted once only)
• Expected frequency in each cell must be ≥ 5 (use Fisher's Exact Test if smaller)
• Categories must be MUTUALLY EXCLUSIVE (one observation fits in only one
category)

Hands-On Classroom Activity

🔍 Activity: 'Do Students Prefer Morning or Evening Study Based on Their Major?'

Instructions (30 minutes): Conduct this with your class in groups of 4-5 students.

1. Collect data: Survey classmates — ask their major (Science/Arts/Commerce) and

preferred study time (Morning/Evening)
2. Create a contingency table: Put majors in rows, study times in columns, fill in counts
3. Calculate expected values: E = (Row Total × Column Total) / Grand Total for each
cell
4. Compute Chi-Square: For each cell, calculate (O-E)²/E, then add all together
5. Find df: (rows-1) × (columns-1)
6. Look up critical value from Chi-Square table at α = 0.05
7. Make your decision and write a one-sentence conclusion

Discussion Questions: Was there a significant association? Does the result surprise you?
What might explain it?

Statistics for Educators • Page 5 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 SECTION
ANOVA
Analysis of Variance — Comparing Means Across Multiple Groups

What is ANOVA?
ANOVA (Analysis of Variance) is a statistical test used to compare the means of THREE or more
groups simultaneously. Instead of running multiple t-tests (which increases error), ANOVA does it all
in one test.

💡 Why Not Just Use Multiple t-Tests?

If we compare 3 groups using t-tests, we need 3 comparisons (A vs B, B vs C, A vs C). Each
test has a 5% chance of a Type I error (false positive). With 3 tests, the combined error rises
to ~14%. ANOVA controls this by testing all groups together — error stays at 5%.

A school principal wants to know which of 3 teaching methods (Traditional, Digital,

ANALOGY Hybrid) produces the best exam scores. Instead of comparing them two at a time
— which is slow and error-prone — ANOVA compares all three at once.

The Core Idea: Variance Decomposition

ANOVA works by splitting the total variation in the data into two parts:

Between-Group Variation Within-Group Variation

Difference BETWEEN the group means Variation WITHIN each group (natural scatter)
If groups are truly different, this will be LARGE This is just random variation — we can't
Example: Method A avg = 75, Method B avg = eliminate it
85, Method C avg = 65 Example: In Method A, scores range from 65-
85

F-ratio = Between-Group Variance / Within-Group Variance

Large F → Groups are different | Small F → Groups are similar

Statistics for Educators • Page 6 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 💡 Think of it This Way
If the signal (between-group differences) is much stronger than the noise (within-group
scatter), then F will be large, and ANOVA will detect a significant effect.

Real-Life Example: Fertilizer Effect on Plant Growth

Scenario: An agronomist tests 3 fertilizer types (A, B, C) on plant height (cm) after 4 weeks. She
uses 5 plants per group.

Plant Fertilizer A Fertilizer B Fertilizer C

1 12 22 17
2 14 25 19
3 11 21 16
4 13 24 18
5 10 23 15
Mean 12 23 17

Grand Mean = (12 + 23 + 17) / 3 = 17.33 cm

ANOVA Calculation Walkthrough

Step 1: Calculate Sum of Squares Between (SSB)

SSB = n × Σ(Group Mean - Grand Mean)²

n = number per group = 5

• Fertilizer A: 5 × (12 - 17.33)² = 5 × 28.41 = 142.05

• Fertilizer B: 5 × (23 - 17.33)² = 5 × 32.15 = 160.75
• Fertilizer C: 5 × (17 - 17.33)² = 5 × 0.11 = 0.55
SSB = 142.05 + 160.75 + 0.55 ≈ 303.33

Step 2: Calculate Sum of Squares Within (SSW)

SSW = Σ(each score - its group mean)²

Statistics for Educators • Page 7 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 Add up squared deviations from each group's mean

• Fertilizer A: (12-12)² + (14-12)² + (11-12)² + (13-12)² + (10-12)² = 0+4+1+1+4 = 10

• Fertilizer B: (22-23)² + (25-23)² + (21-23)² + (24-23)² + (23-23)² = 1+4+4+1+0 = 10
• Fertilizer C: (17-17)² + (19-17)² + (16-17)² + (18-17)² + (15-17)² = 0+4+1+1+4 = 10
SSW = 10 + 10 + 10 = 30

Step 3: The ANOVA Summary Table

Source SS df MS (SS/df) F-ratio

Between Groups 303.33 2 151.67 F = 151.67/2.50 =

60.67

Within Groups 30 12 2.50

Total 333.33 14

F = 60.67 is very large. Critical F at α=0.05, df=(2,12) = 3.89

60.67 > 3.89 → REJECT H₀

DECISION Conclusion: There IS a significant difference in plant heights between
fertilizer types.

Post-Hoc Tests: Finding WHICH Groups Differ

ANOVA only tells us that at least one group is different — not which groups differ. Post-hoc tests
answer this:

Post-Hoc Test When to Use Feature

Tukey's HSD Equal group sizes Most commonly used, controls

family-wise error
Bonferroni Any situation More conservative, adjusts alpha
level
Scheffe Unequal sizes Very conservative, best for complex
contrasts

Statistics for Educators • Page 8 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

Assumptions of ANOVA

⚠️ Check These Before Running ANOVA

• Normality: Data within each group should be approximately normally distributed
• Homogeneity of Variance: All groups should have similar variance (use Levene's
test to check)
• Independence: Observations must be independent of each other
• Random Sampling: Data should be randomly sampled from the population

Hands-On Classroom Activity

🔍 Activity: 'Does Study Duration Affect Quiz Scores?'

Setup: Divide class into 3 groups — Group 1 studies 30 min, Group 2 for 60 min, Group 3 for
90 min. All take the same quiz.

8. Record each student's quiz score (out of 20)

9. Calculate the mean score for each group
10. Construct the ANOVA table (SSB, SSW, df, MS, F)
11. Compare F-calculated to F-critical (from F-table, df = 2, N-3)
12. Write your conclusion — does study time matter?
13. If significant: discuss which pair of groups drives the difference

Extension: What other variables might explain differences in quiz scores? (Prior knowledge,
interest, sleep — discuss confounders!)

Statistics for Educators • Page 9 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 SECTION
P-VALUE
The Probability of Your Results Under the Null Hypothesis

What is a P-Value?
The p-value is one of the most widely used — and misunderstood — concepts in statistics. Here is
the precise, correct definition:

The p-value is the probability of obtaining results AS EXTREME OR MORE

OFFICIAL
DEFINITION
EXTREME than those observed, ASSUMING the null hypothesis (H₀) is
true.

This sounds complex, so let us break it down with a concrete story...

The Coin Flip Story

Scenario: Your friend claims their coin is fair. You flip it 10 times and get 9 Heads. How suspicious
should you be?

• H₀ (Null hypothesis): The coin is fair (P(Heads) = 0.5)

• H₁ (Alternative hypothesis): The coin is unfair (P(Heads) ≠ 0.5)

If the coin were truly fair, getting 9 or 10 heads in 10 flips would happen about 2.1% of the time (p ≈
0.021).

💡 The p-value Interpretation

p = 0.021 means: "IF the coin were fair, there's only a 2.1% chance of seeing 9+ heads."
Since 2.1% is below our threshold of 5% (α = 0.05), we reject the null and conclude the coin
is likely unfair.

Statistics for Educators • Page 10 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

Understanding the Significance Level (α)
The significance level α (alpha) is the threshold we set before collecting data. It represents: 'How
often am I willing to be wrong when I say something is significant?'

Alpha (α) Meaning When Used Risk

α = 0.10 10% false positive rate Exploratory research, High (lenient)

early studies

α = 0.05 5% false positive rate Standard in most social Moderate (common

sciences default)

α = 0.01 1% false positive rate Medical trials, safety Low (strict)

studies

α = 0.001 0.1% false positive rate Physics, genomics studies Very Low (very strict)

The P-Value Decision Framework

p-value Range Decision What it Means

p < 0.001 Highly Significant ✓✓✓ Extremely strong evidence against H₀

p < 0.01 Very Significant ✓✓ Very strong evidence against H₀

p < 0.05 Significant ✓ Strong enough evidence — standard

threshold

0.05 < p < 0.10 Marginal / Borderline Weak evidence; interpret with caution
p ≥ 0.10 Not Significant ✗ Insufficient evidence to reject H₀

Common Misconceptions About P-Values

⚠️ These Misinterpretations Are Extremely Common — Teach These Carefully!

❌ WRONG: "p = 0.03 means there is a 3% chance H₀ is true"

✓ CORRECT: p tells you the probability of seeing this data IF H₀ were true — not the
probability that H₀ is true

❌ WRONG: "p < 0.05 means the result is practically important"

✓ CORRECT: Statistical significance ≠ practical significance. A drug might significantly lower
blood pressure by 0.5 mmHg — statistically significant but clinically useless

Statistics for Educators • Page 11 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 ❌ WRONG: "p > 0.05 means H₀ is proven true / the effect is zero"
✓ CORRECT: Failing to reject H₀ only means we lack evidence against it — absence of
evidence is not evidence of absence

❌ WRONG: "A smaller p-value means a bigger effect"

✓ CORRECT: p-values are affected by sample size. With 1 million subjects, even a trivial
effect can produce p < 0.001

One-Tailed vs Two-Tailed Tests

Type When to Use Example

Two-Tailed When you care if effect is in EITHER Is new drug DIFFERENT from
direction (bigger or smaller) placebo? (could be better OR worse)
One-Tailed (Right) When you care only about effect Is new method BETTER than old
being LARGER method?
One-Tailed (Left) When you care only about effect Does relaxation training REDUCE
being SMALLER anxiety?

Real-World P-Value Examples

🔍 Clinical Trial: Is the New Medicine Better?

• Study: New pain medication vs. placebo in 100 patients

• Result: New drug reduced pain by average 2.3 points more than placebo
• p-value = 0.03
• Interpretation: IF the drug had NO real effect, we'd see this large a difference only
3% of the time
• Decision (α = 0.05): 0.03 < 0.05 → REJECT H₀ → Evidence that drug works
• But: Is 2.3 points on a 100-point scale clinically meaningful? That's a separate
judgment!

🔍 Quality Control: Are Factory Bolts the Right Size?

Statistics for Educators • Page 12 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 • H₀: Mean bolt diameter = 10mm (factory standard)
• Sample of 50 bolts: Mean = 10.2mm
• p-value = 0.08
• Decision (α = 0.05): 0.08 > 0.05 → Fail to Reject H₀
• Interpretation: Not enough evidence to say the machine is miscalibrated — could be
random variation
• Next step: Take a larger sample or investigate further

Hands-On Classroom Activity

🔍 Activity: 'Feeling the p-value — The Dice Experiment'

Goal: Experience what p-values feel like through physical experimentation

14. Each student rolls a die 20 times and records results

15. Calculate observed frequency for each face (1-6)
16. Expected frequency = 20/6 ≈ 3.33 per face
17. Use Chi-Square to test if the die is fair (relates back to Section e!)
18. Report the p-value
19. Class discussion: How many students rejected H₀ at α=0.05? Was it approximately
5% of the class? Why?

Key insight: If H₀ is true (dice are fair), about 5% of students should still get p < 0.05 just by
chance — this demonstrates what Type I error means!

Statistics for Educators • Page 13 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 SECTION
TABLES & CRITICAL VALUES
The Gatekeepers of Statistical Significance

What is a Critical Value?

A critical value is a threshold number obtained from a statistical table that serves as a cut-off point
for making a decision. It is the boundary between the 'rejection region' and the 'non-rejection region'.

Think of a critical value like a passing grade. If a student scores ≥ 50, they pass.
If below 50, they fail. The critical value (50) is the threshold that determines the
ANALOGY
outcome. In statistics, if your test statistic is beyond the critical value, your result
is significant.

The Relationship Between α, p-value, and Critical Value

Concept What It Is How It Is Used

Alpha (α) Your chosen risk level (usually Set BEFORE the study; defines
0.05) where critical value falls
Critical Value The specific threshold number Compare your test statistic
from tables against this number
P-Value Probability computed FROM Compare against alpha to make
your data decision

💡 Two Equivalent Decision Methods

Method 1 (Test Statistic vs Critical Value): If |test statistic| > critical value → Reject H₀
Method 2 (P-Value vs Alpha): If p-value < α → Reject H₀
Both methods ALWAYS give the same answer — they are two sides of the same coin!

How to Read the Chi-Square Table

Navigating the table: Rows represent degrees of freedom (df). Columns represent the significance
level (α). The cell value is the critical value.

Statistics for Educators • Page 14 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

Chi-Square Critical Values Table (Selected Values)

df ↓ α → α = 0.10 α = 0.05 α = 0.025 α = 0.01

1 2.706 3.841 ★ 5.024 6.635

2 4.605 5.991 ★ 7.378 9.210

3 6.251 7.815 ★ 9.348 11.345

4 7.779 9.488 ★ 11.143 13.277

5 9.236 11.070 ★ 12.833 15.086

10 15.987 18.307 ★ 20.483 23.209

20 28.412 31.410 ★ 34.170 37.566

★ = α = 0.05 column (most commonly used). Values highlighted in yellow.

🔍 How to Use: Step-by-Step Table Reading

20. Calculate your test statistic (χ² = 18.18 from our gender/subject example)
21. Calculate df = (rows-1)(cols-1) = (2-1)(2-1) = 1
22. Choose α = 0.05
23. Go to the Chi-Square table: Find row df=1, column α=0.05
24. Critical value = 3.841
25. Compare: 18.18 > 3.841 → Reject H₀

How to Read the F-Distribution Table (For ANOVA)

F-tables are more complex because they have two types of df: df1 (numerator = between groups)
and df2 (denominator = within groups).

F Critical Values at α = 0.05 (Selected Values)

df2 \ df1 df1=1 df1=2 df1=3 df1=4 df1=5

df2=5 6.61 5.79 5.41 5.19 5.05

df2=10 4.96 4.10 3.71 3.48 3.33

df2=12 ★ 4.75 3.89 ★ 3.49 3.26 3.11

Statistics for Educators • Page 15 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 df2=20 4.35 3.49 3.10 2.87 2.71

df2=30 4.17 3.32 2.92 2.69 2.53

★ = Our fertilizer example: df1=2 (between), df2=12 (within). Critical value = 3.89. Our F = 60.67 >>
3.89 → Very significant!

The Complete Decision Framework

Test Find Critical Value Using... Reject H₀ When...

Chi-Square (χ²) df = (r-1)(c-1) and α χ² calculated > χ² critical

ANOVA (F-test) df1 = k-1, df2 = N-k, and α F calculated > F critical
t-test df = n-2 or n1+n2-2, and α |t| calculated > t critical
z-test α only (for large samples) |z| calculated > z critical

Type I and Type II Errors

Choosing a significance level is always a trade-off between two types of errors:

H₀ is Actually TRUE H₀ is Actually FALSE

We REJECT H₀ TYPE I ERROR (α) CORRECT DECISION ✓

False Positive — Detecting an True Positive — We correctly
effect that doesn't exist. detected a real effect (power = 1 -
Controlled by choosing α. β)

We FAIL TO Reject H₀ CORRECT DECISION ✓ TYPE II ERROR (β)

True Negative — We correctly False Negative — Missing a real
found no effect when none exists effect. Reduced by increasing
sample size.

Hands-On Classroom Activity

🔍 Activity: 'Table Treasure Hunt — Master the Reference Tables'

In pairs, answer these questions using only the tables provided in this document:

Statistics for Educators • Page 16 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

 26. What is the critical value for χ² with df=3 at α=0.05?
27. Your Chi-Square result is 8.5. Your df=2. Is this significant at α=0.05?
28. For ANOVA with df1=2, df2=20 at α=0.05, what is the critical F value?
29. You calculated F=3.6 with df1=2, df2=20. Do you reject H₀?
30. A researcher sets α=0.01 for a chi-square test with df=5. What is their critical value?
31. CHALLENGE: You want to be stricter (α=0.01). How does the critical value change
compared to α=0.05? What does this imply about Type I error?

Answer Key: 1) 7.815 2) Yes (8.5 > 5.991) 3) 3.49 4) No (3.6 > 3.49, barely) 5) 15.086
6) Critical value increases → harder to reject H₀ → lower Type I error risk

Statistics for Educators • Page 17 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

QUICK REFERENCE SUMMARY

Chi-Square ANOVA P-Value Critical Values

• Categorical data • Compare 3+ means • p < α → Reject H₀ • From distribution tables

• Tests relationships • F = Between/Within • p ≠ P(H₀ is true) • Depends on df & α
• χ² = Σ(O-E)²/E • One ANOVA = safer • Small p = unusual • Higher α → lower CV
• df=(r-1)(c-1) • Needs post-hoc • Set α before study • 2 methods, same
• Need E ≥ 5 • df: k-1 & N-k • Context matters! answer
• Type I error = α

When to Use Which Test?

Your Data Situation Use This Test Key Output

2 categorical variables, want to Chi-Square (Independence) χ² value + p-value

know if related
Observed counts vs. expected Chi-Square (Goodness of Fit) χ² value + p-value
distribution
3 or more groups, comparing One-Way ANOVA F-ratio + p-value
means
2 factors affecting outcome Two-Way ANOVA 2 F-ratios + interaction
ANOVA significant, want to Post-Hoc (Tukey's, etc.) Pairwise comparisons
know which groups differ
Any test — interpreting the Use p-value vs α Significant or Not
output
Any test — manual comparison Use Critical Value Reject or Fail to Reject

Key Formulas at a Glance

Test Formula Notes
Chi-Square χ² = Σ [ (O-E)² / E ] E = (Row Total × Col Total)
/ Grand Total
Expected Value E = (Row Total × Col Total) Must be ≥ 5 in each cell
/ Grand Total

Chi-Square df df = (rows - 1) × (cols - 1) For contingency tables

Statistics for Educators • Page 18 of 19 • Chi-Square | ANOVA | P-Value | Critical Values

ANOVA F-ratio F = MSB / MSW = (SSB/df1) / df1=k-1, df2=N-k
(SSW/df2)
SSB n × Σ(Group Mean - Grand Between-group sum of squares
Mean)²
SSW Σ(each score - its group Within-group sum of squares
mean)²

End of Study Material

Statistics for Educators • Chi-Square | ANOVA | P-Value | Critical Values

Statistics for Educators • Page 19 of 19 • Chi-Square | ANOVA | P-Value | Critical Values