Inferential Statistics • Lesson 12

Null & Alternative Hypotheses

Learn how statistical hypotheses turn a real-world claim into a testable statement about a population.

H₀Hₐ / H₁Two-Sided TestsOne-Sided Tests

What You Will Learn

1

Understand the purpose of statistical hypotheses.

2

Distinguish between the null and alternative hypotheses.

3

Write hypotheses using population parameters.

4

Identify two-sided, left-tailed, and right-tailed alternatives.

5

Translate business questions into statistical hypotheses.

6

Avoid common mistakes when writing H₀ and Hₐ.

Why Do We Need Hypotheses?

In real-world analytics, we often want to test a claim. For example, a company may claim that its average delivery time is 30 minutes.

We usually cannot measure every delivery in the population. Instead, we collect a sample and use statistical inference to evaluate whether the sample provides enough evidence against the claim.

Claim

Population statement

Sample

Collect evidence

Decision

Evaluate the claim

STEP 1

Start With a Population Claim

Suppose a food-delivery company claims:

Company claim

μ = 30 minutes

Here, μ represents the true population mean delivery time.

We can use hypotheses to formally represent this claim and the possibility that the claim is not correct.

STEP 2

The Two Hypotheses

NULL HYPOTHESIS

H₀

μ = 30

The null hypothesis represents the reference claim or baseline assumption that we test against.

ALTERNATIVE HYPOTHESIS

Hₐ

μ ≠ 30

The alternative hypothesis represents the pattern we want to find evidence for instead of the null hypothesis.

STEP 3

Why Is Equality Usually in H₀?

In the standard hypothesis-testing setup, the null hypothesis contains the equality because it provides a specific reference value for calculating how unusual the sample result would be.

H₀

μ = 30

Hₐ

μ ≠ 30

Evidence

Sample data

STEP 4

Three Common Forms of Hₐ

Test TypeAlternativeMeaning
Two-sidedμ ≠ μ₀The population value may be higher or lower.
Right-tailedμ > μ₀The population value is greater.
Left-tailedμ < μ₀The population value is smaller.

Two-Sided Test

Use a two-sided alternative when the question is whether the population parameter is different from the reference value in either direction.

H₀: μ = 50

Hₐ: μ ≠ 50

Example question: “Has the average delivery time changed from 50 minutes?”

Right-Tailed Test

Use a right-tailed alternative when the research question is specifically asking whether the population value is greater.

H₀: μ = 50

Hₐ: μ > 50

Example question: “Is the average delivery time greater than 50 minutes?”

Left-Tailed Test

Use a left-tailed alternative when the research question is specifically asking whether the population value is smaller.

H₀: μ = 50

Hₐ: μ < 50

Example question: “Has the average delivery time fallen below 50 minutes?”

STEP 5

Translate Business Questions Into Hypotheses

Question 1: Is the average customer spending different from ₹2,000?

H₀: μ = ₹2,000

Hₐ: μ ≠ ₹2,000

Question 2: Is the average delivery time greater than 30 minutes?

H₀: μ = 30

Hₐ: μ > 30

Question 3: Is the defect rate below 5%?

H₀: p = 0.05

Hₐ: p < 0.05

Population Parameter vs Sample Statistic

Hypotheses describe the population parameter. The sample statistic is then used as evidence when testing the hypothesis.

ConceptExample
Population meanμ
Sample meanx̄
Population proportionp
Sample proportionp̂

Key Idea

Think of H₀ as the Reference Model

The hypothesis test asks whether the observed sample evidence would be sufficiently unusual if the null hypothesis were the reference assumption.

H₀

Reference assumption

Sample

Observed evidence

Test

Quantifies how unusual the evidence is

Common Mistakes

❌ Putting the equality only in Hₐ

In the standard setup, equality is placed in H₀.

❌ Using the sample statistic as the hypothesis

Hypotheses are statements about population parameters, not simply observed sample values.

❌ Choosing the tail after seeing the data

The direction of the alternative should come from the research question and be specified before analyzing the result.

❌ Saying “fail to reject” means H₀ is proven

Failing to reject H₀ means the evidence was not sufficient to reject it under the chosen testing procedure.

PRACTICE

Test Your Understanding

CHECK YOUR UNDERSTANDING

Which hypothesis represents the reference claim being tested?

CHECK YOUR UNDERSTANDING

If the research question asks whether μ is different from 100, what should Hₐ be?

CHECK YOUR UNDERSTANDING

Which alternative represents a right-tailed test?

CHECK YOUR UNDERSTANDING

A company wants to know whether its average delivery time is below 30 minutes. Which Hₐ is appropriate?

CHECK YOUR UNDERSTANDING

Which symbol normally represents the population proportion?

CHECK YOUR UNDERSTANDING

If H₀: μ = 500 and Hₐ: μ ≠ 500, what type of test is this?

Fill in the Blank

Fill in the Blank

The null hypothesis is commonly written as H__.

Fill in the Blank

The alternative hypothesis is commonly written as H__.

Fill in the Blank

If Hₐ: μ > 40, the test is ______-tailed.

Fill in the Blank

If Hₐ: μ < 40, the test is ______-tailed.

Analytics Challenge

Turn a Business Question Into Hypotheses

An online store claims that its average order value is ₹2,500. An analyst wants to determine whether the true average order value is different from ₹2,500.

Your task:

Write H₀ and Hₐ.

H₀: μ = ₹2,500

Hₐ: μ ≠ ₹2,500

Because the question asks whether the value is “different,” the alternative must allow both higher and lower values.

Final Challenge

Choose the Correct Alternative

A website team wants to know whether its average page-load time has increased above the target of 2.5 seconds.

What should the alternative hypothesis be?

Hₐ: μ > 2.5 seconds

The word “increased” tells us that the research question is directional, so this is a right-tailed alternative.

Lesson 12 Summary

✓

H₀ is the null hypothesis and acts as the reference assumption.

✓

Hₐ represents the alternative claim we want evidence for.

✓

Hypotheses describe population parameters such as μ or p.

✓

Hₐ: μ ≠ μ₀ creates a two-sided test.

✓

Hₐ: μ > μ₀ creates a right-tailed test.

✓

Hₐ: μ < μ₀ creates a left-tailed test.

✓

The direction of the test should come from the research question, not from the observed result.

✓

Failing to reject H₀ does not mean H₀ has been proven true.