Alternative Hypothesis

Definition

Basic Concept

The alternative hypothesis (denoted H_a or H₁) is a statement that contradicts the null hypothesis (H₀). It represents what the researcher aims to support through data evidence.

Purpose

To provide a specific claim about a population parameter that differs from the null hypothesis, enabling statistical testing to determine if observed data favor this claim.

Notation

Commonly denoted as H_a or H₁. Example: H₀: μ = 10, H_a: μ ≠ 10.

Role in Hypothesis Testing

Testing Framework

Hypothesis testing evaluates evidence against H₀. The alternative hypothesis represents the rival claim tested indirectly via rejecting or failing to reject H₀.

Decision Making

Data either provide sufficient evidence to reject H₀ in favor of H_a or fail to do so, leaving H₀ un-rejected.

Inference Basis

Supports inferential conclusions about population parameters based on sample data and probabilistic thresholds.

Formulation

Parameter Specification

H_a specifies an inequality or difference concerning a population parameter (mean, proportion, variance, etc.).

Examples

H_a: μ > μ₀, H_a: p < 0.5, H_a: σ² ≠ 4.

Context Dependence

Formulation depends on research question, measurement scale, and hypothesis directionality.

Types of Alternative Hypotheses

One-Tailed (Directional)

Specifies a direction: greater than (>) or less than (<) a parameter value. Example: H_a: μ > 50.

Two-Tailed (Non-Directional)

Specifies inequality without direction: ≠. Example: H_a: μ ≠ 100.

Composite vs Simple

Composite: involves range of values. Simple: specifies a single value.

Relationship to Null Hypothesis

Mutually Exclusive

H_a and H₀ cannot both be true simultaneously.

Exhaustive

Together, H₀ and H_a cover all possible values of the parameter.

Testing Contrast

Hypothesis test outcome rejects or fails to reject H₀, indirectly supporting H_a.

Test Statistics and Decision Rules

Test Statistic Calculation

Computed from sample data; measures how much data deviate from H₀ assumption.

Decision Rule

Reject H₀ if test statistic falls in critical region defined by H_a.

Example Table

One-tailed vs Two-tailed Tests

Directionality

One-tailed tests assess deviation in one direction; two-tailed assess deviation in both.

Critical Regions

One-tailed: single rejection region. Two-tailed: two rejection regions at both ends of distribution.

Examples

One-tailed: H_a: p < 0.05. Two-tailed: H_a: p ≠ 0.05.

P-value Interpretation

Definition

Probability of observing test statistic as extreme or more extreme than actual, assuming H₀ true.

Comparison to Significance Level

Reject H₀ if p-value ≤ α (significance level), supporting H_a.

Misinterpretations

Does not measure probability H₀ is true or false; indicates data extremity under H₀.

Significance Levels

Definition

α: threshold for Type I error probability; commonly 0.05, 0.01, or 0.10.

Role in Testing

Determines rejection region boundaries for test statistic based on H_a type.

Trade-offs

Lower α reduces false positives but increases false negatives (Type II error).

Examples

Mean Difference Test

H₀: μ = 50, H_a: μ ≠ 50. Data: sample mean = 54, standard deviation = 10, n = 30.

Proportion Test

H₀: p = 0.5, H_a: p > 0.5. Sample proportion = 0.6, n = 100.

Variance Test

H₀: σ² = 16, H_a: σ² < 16. Sample variance = 12, n = 25.

Example: One-sample t-test for meanGiven: H0: μ = 50 Ha: μ ≠ 50 Sample mean (x̄) = 54 Sample std dev (s) = 10 Sample size (n) = 30Test statistic: t = (x̄ - μ0) / (s / √n) t = (54 - 50) / (10 / √30) ≈ 2.19Decision: Compare t to critical t for df=29 at α=0.05 (two-tailed) If |t| > t_critical, reject H0 in favor of Ha

Common Misconceptions

Rejecting H₀ Proves H_a

Rejecting H₀ supports but does not prove H_a. Statistical evidence is probabilistic.

Failing to Reject H₀ Proves It True

Failing to reject H₀ indicates insufficient evidence, not proof of truth.

Alternative Hypothesis is Always True if H₀ is False

Some tests may have multiple alternative hypotheses or inconclusive results.

Importance in Statistics

Foundation for Inferential Statistics

Defines the research hypothesis tested through data; essential for scientific conclusions.

Guides Experimental Design

Formulates expected effects, influencing sample size and analysis methods.

Supports Decision Making

Enables objective assessment of claims using probability theory and observed data.

References

• Casella, G., & Berger, R. L. Statistical Inference, 2nd ed., Duxbury, 2002, pp. 125-180.
• Mendenhall, W., Beaver, R. J., & Beaver, B. M. Introduction to Probability and Statistics, 14th ed., Cengage, 2012, pp. 320-365.
• Lehmann, E. L., & Romano, J. P. Testing Statistical Hypotheses, 3rd ed., Springer, 2005, pp. 45-95.
• Wasserman, L. All of Statistics: A Concise Course in Statistical Inference, Springer, 2004, pp. 110-150.
• Devore, J. L. Probability and Statistics for Engineering and the Sciences, 9th ed., Cengage, 2011, pp. 250-290.