In statistics and data science, hypothesis testing is a core method used to make decisions based on data. This process are two competing statements: the Null Hypothesis and the Alternative Hypothesis. Understanding the difference between them is essential for correct analysis and interpretation.
What is a Null Hypothesis (H0)?
The Null Hypothesis is the starting assumption in hypothesis testing. It states that there is no effect, no difference or no relationship between variables.
- Assumes the current situation remains unchanged
- States no change or no impact
- Directly tested using sample data
- Written using equality symbols ( =, ≥, ≤ )
Example:
H₀: The average salary of data scientists is ₹10 LPA.
H₀: There is no difference between Model A and Model B accuracy.
What is an Alternative Hypothesis (H₁ or Hₐ)?
The Alternative Hypothesis is the statement that challenges the null hypothesis. It suggests that an effect, a difference or a relationship does exist.
- Represents the research or study claim
- Indicates that a change or effect is present
- Accepted only when the null hypothesis (H₀) is rejected
- Written using inequality symbols ( ≠, >, < )
Example:
H₁: The average salary of data scientists is not ₹10 LPA.
H₁: Model A performs better than Model B.
Difference between the null hypothesis and alternate hypothesis
| Null Hypothesis (H₀) | Alternative Hypothesis (H₁ / Hₐ) |
|---|---|
| States that no relationship exists between variables | States that a relationship exists between variables |
| Assumes no effect or no change | Assumes an effect or change is present |
| Considered the default assumption | Represents the research claim |
| Researchers try to reject this hypothesis | Researchers aim to support/accept this hypothesis |
| If accepted (not rejected), researchers may need to reconsider their assumptions | If accepted, researchers support their original claim |
| No observable impact on the output | Observable impact on the output |
| Tested directly using statistical tests | Supported indirectly by rejecting H₀ |
| Denoted by H₀ | Denoted by H₁ or Hₐ |
| Accepted when p-value > significance level (α) | Accepted when p-value < significance level (α) |
| Maintained when there is insufficient evidence | Accepted when there is strong evidence |