Last modified: February 05, 2025

This article is written in: 🇺🇸

Null Hypotheses and Alternative Hypotheses

Statistical hypothesis testing is a fundamental method used in research to make inferences about populations based on sample data. Understanding the concepts of null and alternative hypotheses, as well as how to calculate and interpret p-values, is crucial for conducting robust and meaningful analyses. This section delves into these concepts, providing a comprehensive overview to enhance your statistical toolkit.

After reading this material you should be able to tell:

What is Null Hypothesis?

In statistical hypothesis testing, we construct two opposing hypotheses to make inferences about a population parameter based on sample data:

  1. The null hypothesis (H0) represents the assumption that there is no effect, no difference, or no change in the population. It serves as the default assumption in hypothesis testing. For example, if we are testing a new drug, the null hypothesis might state that the drug has no effect on patients compared to a placebo.
  2. The alternative hypothesis (H1 or Ha) challenges the null hypothesis by suggesting that there is an effect, a difference, or a change in the population. Continuing with the previous example, the alternative hypothesis would assert that the drug does have an effect on patients.

Confused now? That's understandable. Choosing a null hypothesis is often guided by heuristics, and there is no strictly defined method for selecting it. By examining numerous examples, you will begin to develop an intuition for what researchers typically mean by the null hypothesis and how they choose it. This practical exposure is important for truly mastering the concept.

The relationship between the null and alternative hypotheses is foundational to the scientific method, as it allows researchers to test assumptions and determine whether an observed effect is genuine or simply due to random chance. Null hypothesis significance testing (NHST) is used across various academic disciplines, including medicine, psychology, neuroscience, genetics, economics, linguistics, and more.

Now, pay close attention! If that wasn't confusing enough, there's more to the story. Another concept is closely related, and together they form the foundation of NHST. As we already learned the primary goal of NHST is to use statistical evidence from sample data to decide whether to reject the null hypothesis in favor of the alternative hypothesis. This decision is based on the p-value, which quantifies the probability of obtaining results at least as extreme as those observed, assuming that the null hypothesis is true.

Null Hypothesis Testing Illustration

The plot illustrates the distribution of a test statistic under the assumption that the null hypothesis H0 is true. Here's a detailed explanation of the components:

The Language of Hypothesis Testing

Understanding P-Values Through an Analogy

Imagine you're a detective tasked with determining whether a suspect is guilty of a crime. In this scenario, the null hypothesis (H0) is that the suspect is innocent, while the alternative hypothesis (Ha) is that the suspect is guilty.

As you gather evidence from the crime scene—such as fingerprints, DNA samples, and eyewitness accounts—you assess how likely it is to find this evidence if the suspect were truly innocent. The p-value represents the probability of obtaining evidence as strong as what you have (or stronger) under the assumption that the suspect is innocent.

In a one-tailed test, you might only consider evidence that points towards guilt, calculating the probability of finding such incriminating evidence. This means you're looking at one direction of the distribution of possible evidence. Conversely, in a two-tailed test, you also account for evidence that could exonerate the suspect, considering both directions—whether the evidence is strongly incriminating or strongly exonerating.

If the p-value is low, say below a 5% significance level, it suggests that the observed evidence is unlikely to have occurred by chance if the suspect were innocent. This would lead you to reject the null hypothesis and conclude that there is sufficient evidence to support the suspect's guilt. On the other hand, a high p-value would indicate that the evidence is not strong enough to rule out innocence, and you would fail to reject the null hypothesis.

Interpretation of the P-value

The p-value is the probability of obtaining a test statistic at least as extreme as the one observed, assuming that the null hypothesis H0 is true.

Example: Coin Toss Experiment

Statistical Test

We use the binomial test to calculate the p-value for the observed outcome.

I. Compute the probability of observing exactly 4 heads under H0:

P(X=4)=(104)(0.5)4(0.5)6=210×0.0625×0.0156251=0.2051

II. Compute the probabilities for more extreme outcomes (0 to 4 heads and 6 to 10 heads) since it's a two-tailed test.

III. Total P-value:

p-value=2×(P(X4))=2×(k=04P(X=k))

IV. Sum of Probabilities:

$$ \begin{align} P(X = 0) &= 0.00098 \\ P(X = 1) &= 0.00977 \\ P(X = 2) &= 0.04395 \\ P(X = 3) &= 0.11719 \\ P(X = 4) &= 0.20508 \\ \sum_{k=0}^{4} P(X = k) &= 0.37697 \end{align} $$

V. Compute P-value:

p-value=2×0.37697=0.75394

Decision

Additional Considerations

Table of Contents

    Null Hypotheses and Alternative Hypotheses
    1. What is Null Hypothesis?
    2. The Language of Hypothesis Testing
    3. Understanding P-Values Through an Analogy
    4. Interpretation of the P-value
    5. Example: Coin Toss Experiment
      1. Statistical Test
      2. Decision
    6. Additional Considerations