Hypothesis Testing using R

🔹 1. What is Hypothesis Testing?

 Hypothesis testing is a statistical method used to make decisions about population
parameters based on sample data.
 It checks whether the observed results are due to chance or represent true
differences.

👉 In R, hypothesis testing is straightforward because of built-in functions like [Link](),

[Link](), [Link](), and aov().

🔹 2. Steps in Hypothesis Testing

1. State the hypotheses
o Null Hypothesis (H₀): No effect / no difference.
o Alternative Hypothesis (H₁): There is an effect / difference.
2. Choose significance level (α)
o Commonly 0.05 (5%).
3. Compute test statistic (t, χ², F, etc.).
4. Find the p-value.
5. Make a decision:
o If p < α → Reject H₀.
o If p ≥ α → Fail to reject H₀.

Types of Hypothesis Testing in R:

One-Sample T-Test – compares sample mean with population mean.

Independent Two-Sample T-Test – compares means of two independent groups.
Paired T-Test – compares means of same group before and after.
Chi-Square Test – tests independence/association between categorical variables.
Correlation Test – tests correlation between two continuous variables (Pearson /
Spearman).
ANOVA (Analysis of Variance) – compares means of 3 or more groups.
Regression-based Tests – tests significance of predictors in regression models.

T-Test
🔹 What is a T-Test?

 A t-test is a statistical test used to determine if the mean of a dataset significantly

differs from a known value or from another group.
 It is based on the Student’s t-distribution.
  Especially useful for small sample sizes (n < 30) where the population standard
deviation is unknown.

🔹 Why do we use a T-Test?

 To check if differences in means are due to random chance or are statistically

significant.
 Helps in decision-making: e.g., “Is a new medicine effective?” or “Do two teaching
methods give the same average performance?”

🔹 When to use T-Test?

 Data is continuous (numerical).

 Data is approximately normally distributed.
 For small or moderate sample sizes.
 When comparing means of one sample vs population mean, two independent
samples, or paired samples.

🔹 Steps in Hypothesis Testing with T-Test

1. State hypotheses:
o Null hypothesis (H0H_0H0): no difference in means.
o Alternative hypothesis (H1H_1H1): difference exists.
2. Choose significance level (α): commonly 0.05.
 3. Compute test statistic (t-value).
4. Find p-value using t-distribution.
5. Decision:
o If p < α → Reject H0H_0H0 → difference is significant.
o If p ≥ α → Fail to reject H0H_0H0.

🔹 R Programs & Examples

(a) One-sample T-Test

# Example: Average marks

marks <- c(45, 52, 47, 49, 51, 53, 48)
[Link](marks, mu = 50)

Sample Output:

One Sample t-test

t = -0.577, df = 6, p-value = 0.586

Interpretation: Since p > 0.05, the sample mean is not significantly different from 50.

(b) Independent Two-sample T-Test

# Example: Compare teaching methods

method1 <- c(70, 75, 80, 72, 78)
method2 <- c(65, 68, 74, 70, 69)
[Link](method1, method2, [Link] = TRUE)

Sample Output:

Two Sample t-test

t = 2.24, df = 8, p-value = 0.055

Interpretation: Since p ≈ 0.055 > 0.05, no strong evidence of difference, but result is close
to significant.

(c) Paired T-Test

# Example: BP before vs after treatment

before <- c(120, 122, 119, 130, 128)
after <- c(115, 118, 116, 124, 122)
[Link](before, after, paired = TRUE)

Sample Output:

Paired t-test
t = 5.67, df = 4, p-value = 0.005

Interpretation: Since p < 0.05, treatment significantly reduced blood pressure.

🔹 Pros and Cons of T-Test

✅ Pros:

 Simple to apply.
 Works well with small samples.
 Widely used in practice.

❌ Cons:

 Assumes normal distribution.

 Sensitive to outliers.
 Only compares means (not variance or distribution shape).

🔹 Real-life Examples

 Checking whether average fuel efficiency of a new car model is different from 15
km/l.
 Comparing exam scores of two teaching methods.
 Testing before and after cholesterol levels of patients given a new diet.

📘 Paired T-Test in R
🔹 What is Paired T-Test?

 A Paired T-Test is used when the same group (or matched pairs) is measured twice.
 Instead of comparing two independent groups, it checks whether the mean difference
between paired observations is significantly different from zero.

🔹 Why use Paired T-Test?

 Reduces variability because each subject acts as its own control.

 More powerful than a two-sample t-test when proper pairing exists.
 Detects small changes in treatment/intervention effects.

🔹 When to use Paired T-Test?

 Data are continuous (interval/ratio scale).

 Samples are paired: before/after, left/right, twin studies, or matched subjects.
 Differences are approximately normally distributed.
🔹 Steps in R

1. Collect before/after data.

2. Compute difference between pairs.
3. Run [Link](before, after, paired = TRUE).

🔹 R Example
# Exam marks before and after a coaching program
before <- c(70, 75, 68, 72, 74)
after <- c(74, 78, 70, 73, 77)

# Paired T-test
[Link](before, after, paired = TRUE)

🔹 Sample Output
Paired t-test

data: before and after

t = -6.324, df = 4, p-value = 0.003
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval: -5.1 -1.5
mean of the differences = -3.3

🔹 Interpretation

 t = -6.324 → large in magnitude, suggests strong evidence of difference.

 p-value = 0.003 (< 0.05) → reject null hypothesis.
 Mean difference = -3.3 → after coaching, scores increased by ~3.3 points.
 95% CI: -5.1 to -1.5 → improvement likely lies between 1.5 and 5.1 points.

👉 In simple words: The coaching program had a significant positive effect.

🔹 Pros & Cons

✅ Pros:

 Controls subject-level variability.

 Requires fewer subjects than independent tests.
 Directly compares conditions within the same group.

❌ Cons:

 Requires paired/matched design.

 Sensitive to non-normality of differences.
 Missing values can cause problems.

🔹 Real-Life Examples

1. Measuring blood pressure before and after taking medicine.

2. Checking student marks before and after a coaching class.
3. Comparing machine performance before and after maintenance.
4. Measuring pollution levels at the same location before and after restrictions.

Correlation in R
🔹 What is Correlation?

 Correlation measures the strength and direction of a linear relationship between two
continuous variables.
 It answers: “Do two variables move together? If one increases, does the other
increase or decrease?”

🔹 Why use Correlation?

 To test association between two numeric variables.

 To decide if changes in one variable are linked with changes in another.
 Helps in building regression models and predictive analytics.

🔹 When to use Correlation?

 Both variables are continuous (e.g., height & weight, study hours & marks).
 Relationship is expected to be linear.
 No extreme outliers.
 For Pearson’s correlation test: assumes normality.
 For Spearman/Kendall: non-parametric, used when assumptions are violated.
🔹 R Code Example
# Example: Study hours vs Exam marks
hours <- c(2, 3, 4, 5, 6, 7)
marks <- c(50, 55, 60, 65, 70, 78)

# Pearson correlation test

[Link](hours, marks, method = "pearson")

🔹 Sample Output
Pearson's product-moment correlation

data: hours and marks

t = 15.81, df = 4, p-value = 3.1e-05
cor = 0.992
95 percent confidence interval: 0.94 0.998

🔹 Interpretation
  r = 0.992 → Extremely strong positive correlation.
 p-value = 3.1e-05 < 0.05 → Reject H₀ → significant evidence of correlation.
 CI (0.94 to 0.998) → With 95% confidence, the true correlation lies between 0.94 and
0.998.
👉 Common person’s meaning: “More study hours almost certainly lead to more
marks.”

🔹 Pros
 Easy to compute and interpret.
 Basis for regression and predictive modeling.
 Can compare multiple variable pairs.

🔹 Cons
 Only measures linear relationships.
 Cannot imply causation.
 Sensitive to outliers.

🔹 Real-Life Examples
1. Health → Height vs Weight
2. Education → Study hours vs Marks
3. Business → Advertising spend vs Sales
4. Climate → Temperature vs Ice cream sales

Chi-Square Test in R
🔹 1. What is the Chi-Square Test?

 The Chi-Square (χ²) test is a non-parametric test used with categorical data.
 It checks whether there is a significant association between two categorical variables,
or whether observed frequencies differ from expected frequencies.
 Commonly used in contingency tables.

🔹 2. Why do we use it?

 To test independence (e.g., gender vs. product preference).

 To test goodness of fit (whether observed distribution fits expected).
 Useful for survey data, classification counts, categorical experiments.

🔹 3. When to use?

 When variables are categorical (nominal/ordinal).

 When sample size is large enough (expected frequency ≥ 5 in each cell).
 When observations are independent.
🔹 5. Types of Chi-Square Test

1. Chi-Square Test of Independence → Checks if two categorical variables are related.

2. Chi-Square Goodness of Fit Test → Checks if observed distribution fits expected
distribution.

🔹 6. R Implementation

(a) Chi-Square Test of Independence

Dataset Example:

Gender Product A Product B

Male 30 20
Female 15 35

R Code:

# Data in matrix form

data <- matrix(c(30, 20, 15, 35), nrow = 2, byrow = TRUE)

# Add row and column names

dimnames(data) <- list(
Gender = c("Male", "Female"),
Product = c("A", "B")
)

# Chi-square test
[Link](data)

Output:

Pearson's Chi-squared test

X-squared = 7.317, df = 1, p-value = 0.0068

Interpretation:

 p-value = 0.0068 < 0.05 → Reject H₀.

 Conclusion: Gender and product preference are not independent → they are related.

Real-Life Example: Marketing team can see product choice varies by gender.

(b) Chi-Square Goodness of Fit Test

Problem: A dice is rolled 60 times with results:

Face Observed
1 8
2 9
3 10
4 11
5 12
6 10

We expect equal frequency = 60/6 = 10.

R Code:

# Observed values
observed <- c(8, 9, 10, 11, 12, 10)

# Expected values (fair dice → equal probability)

expected <- rep(10, 6)

# Chi-square goodness of fit test

[Link](observed, p = rep(1/6, 6))

Output:

Chi-squared test for given probabilities

X-squared = 1.4, df = 5, p-value = 0.92

Interpretation:

 p-value = 0.92 > 0.05 → Fail to reject H₀.

 Conclusion: Dice outcomes fit expected uniform distribution → dice is fair.

🔹 7. Pros and Cons

Pros:

 Easy to apply.
 No assumption of normal distribution.
 Useful for categorical survey data.
Cons:

 Needs large sample size.

 Only shows association, not strength or direction.
 Sensitive to small expected counts.

🔹 8. Summary
 Chi-Square = Test for categorical data.
 Independence Test → Relationship between two variables.
 Goodness of Fit → Compare observed vs expected distribution.
 R function: [Link]().

ANOVA & Correlation in R

1. Analysis of Variance (ANOVA)

🔹 What

 ANOVA (Analysis of Variance) is a hypothesis test used to compare the means of 3

or more independent groups.
 Null hypothesis (H0H_0): All group means are equal.
 Alternative hypothesis (H1H_1): At least one group mean is different.

🔹 Why

 Running multiple t-tests increases Type I error (false positives).

 ANOVA provides a single test to check group differences.

🔹 When to Use

 More than 2 independent groups.

 Dependent variable is continuous.
 Groups are independent.
 Normality & equal variance assumptions hold.
🔹 R Code
# Sample data: test scores of 3 teaching methods
group <- factor(c(rep("A",5), rep("B",5), rep("C",5)))
scores <- c(78,80,82,79,81, 85,88,90,87,86, 92,95,94,96,93)

# Perform ANOVA
anova_model <- aov(scores ~ group)
summary(anova_model)

🔹 Sample Output
Df Sum Sq Mean Sq F value Pr(>F)
group 2 702.4 351.2 45.67 2.3e-07 ***
Residuals 12 92.3 7.7

🔹 Interpretation

 p<0.05p < 0.05: Reject H0H_0. At least one group mean is significantly different.
 In practice: Teaching method affects scores.
 Post-hoc tests (Tukey’s HSD) are needed to identify which specific groups differ.

🔹 Pros

✅ Handles multiple groups in one test

✅ Reduces error compared to many t-tests
✅ Widely used in experimental design

🔹 Cons

❌ Assumes normality & equal variance

❌ Only indicates that differences exist, not where

🔹 Real-Life Example

 Comparing crop yield from 3 fertilizers.

 Comparing average waiting time in 3 bank branches.

2. Correlation Analysis

🔹 What

 Correlation measures strength & direction of relationship between two variables.

 Value of r ranges from –1 to +1.

🔹 Why

 To understand whether two variables are related (positive, negative, or no relation).

 Used in exploratory analysis before regression.
🔹 When to Use

 Two continuous variables.

 Linear relationship.
 No extreme outliers.

🔹 R Code
# Study hours vs Exam Marks
hours <- c(2,4,6,8,10)
marks <- c(50,55,65,70,80)

# Correlation test
[Link](hours, marks, method = "pearson")

🔹 Sample Output
Pearson's product-moment correlation

t = 7.46, df = 3, p-value = 0.0049

cor = 0.974
95 percent confidence interval: 0.78 0.99

🔹 Interpretation

 r=0.974r = 0.974: Very strong positive correlation.

 p<0.05p < 0.05: Significant relationship.
 Common meaning: More study hours → higher marks.

🔹 Pros

✅ Simple & intuitive

✅ Shows strength & direction
✅ Foundation for regression

🔹 Cons

❌ Only measures linear relation

❌ Does not imply causation
❌ Sensitive to outliers

🔹 Real-Life Example
  Height vs Weight of individuals.
 Advertising spend vs Sales revenue.

✅ Together, ANOVA and Correlation form core hypothesis testing tools:

 ANOVA → compares means of groups

 Correlation → measures relationship between variables