Hypothesis test for the proportion in a binomial distribution

Welcome to Hypothesis Testing!

Ever wondered if a "lucky" coin is actually biased, or if a new medicine really works better than the old one? Hypothesis testing is the mathematical way of playing detective. In this chapter, we will learn how to use the Binomial Distribution to decide whether a claim about a proportion is likely to be true or if a result happened just by random chance.

Don't worry if this seems a bit "wordy" at first—once you grasp the logic, the math is very straightforward!

1. The "Big Idea": Logic of the Test

Imagine your friend claims they can predict the outcome of a coin toss 80% of the time. You watch them try 10 times, and they get 9 right. Is that just luck, or are they actually "psychic"?

In a hypothesis test, we always start by assuming nothing has changed (the coin is fair). We then calculate how likely it is to get a result as extreme as 9 out of 10. If that probability is tiny, we "reject" the idea that it was just luck.

Key Terms You Need to Know

Null Hypothesis (\(H_0\)): This is the "status quo." We assume the proportion \(p\) is exactly what it's always been.
Example: \(H_0: p = 0.5\) (The coin is fair).

Alternative Hypothesis (\(H_1\)): This is what we are actually testing for. It’s the "suspicion" that the proportion has changed.
Example: \(H_1: p > 0.5\) (The coin is biased towards heads).

Test Statistic: This is the actual number of "successes" we observe in our sample (e.g., the 9 heads your friend flipped).

Significance Level (\(\alpha\)): This is the "bar" we set for proof. Usually, it's 5% (0.05). If the probability of our result is less than this, we say it's "statistically significant."

Quick Review: Hypothesis testing is like a court trial. The defendant is innocent until proven guilty. In math, \(H_0\) is "innocent" (nothing changed), and we need strong evidence to prove "guilt" (\(H_1\)).

2. Setting up the Hypotheses

When writing hypotheses for a Binomial proportion \(p\), you must always use the parameter \(p\) and state what it represents in words.

There are three types of tests you might perform:

1. One-tailed (Upper): Testing if the proportion has increased. \(H_1: p > \text{value}\).
2. One-tailed (Lower): Testing if the proportion has decreased. \(H_1: p < \text{value}\).
3. Two-tailed: Testing if the proportion has changed in either direction. \(H_1: p \neq \text{value}\).

Common Mistake: Never use the sample result in your hypotheses. For example, if you see 8/10 successes, do not write \(H_0: p = 0.8\). The hypotheses are about the whole population, not just your small sample.

3. The Step-by-Step Testing Process

Follow these 5 steps for every question:

Step 1: State your Hypotheses.
Define \(p\) (e.g., "where \(p\) is the probability of a seed germinating").
Write \(H_0: p = \dots\) and \(H_1: p < \dots\) (or \(>\) or \(\neq\)).

Step 2: Identify the Model.
State the distribution under the Null Hypothesis: \(X \sim B(n, p)\).
\(n\) is your sample size, and \(p\) is the value from \(H_0\).

Step 3: Calculate the p-value.
This is the probability of getting your observed result or more extreme.
- If testing \(p > \dots\), calculate \(P(X \ge \text{observed})\).
- If testing \(p < \dots\), calculate \(P(X \le \text{observed})\).

Step 4: Compare to the Significance Level.
If p-value \(\le\) Significance Level, we Reject \(H_0\).
If p-value \(>\) Significance Level, we Fail to Reject \(H_0\).

Step 5: Conclude in Context.
Use phrases like: "There is sufficient evidence at the 5% level to suggest the proportion of... has increased."

Did you know? We never say we "accept" \(H_0\). We just say there isn't enough evidence to reject it. It’s like a jury saying "Not Guilty"—it doesn't necessarily mean "Innocent," just that there wasn't enough proof!

4. Critical Regions and Critical Values

Sometimes, instead of a p-value, you are asked for the Critical Region. This is the range of values for which you would reject \(H_0\).

Critical Value: The first value that falls inside the rejection zone.
Acceptance Region: The values where we stick with \(H_0\).

Important OCR Rule: For a specified significance level (e.g., 5%), the probability of the test statistic being in the rejection region must be as close as possible to the significance level without exceeding it.

The Two-Tailed Twist

In a two-tailed test (\(H_1: p \neq \dots\)), the "suspicion" is just that things have changed. We split the significance level in half.
Example: At a 5% significance level, we look for 2.5% at the bottom end AND 2.5% at the top end.

Memory Aid: Two tails = Two directions = Two halves of the %.

5. Understanding the Significance Level

What does a "5% significance level" actually mean?
It is the probability of incorrectly rejecting the Null Hypothesis. In other words, there is a 5% chance that we conclude something has changed when, in reality, the result was just a rare piece of luck.

Quick Summary Table:
- p-value is small (\(<\alpha\)): Result is rare \(\rightarrow\) Evidence against \(H_0\) \(\rightarrow\) Reject \(H_0\).
- p-value is large (\(>\alpha\)): Result is common \(\rightarrow\) No evidence against \(H_0\) \(\rightarrow\) Do not reject \(H_0\).

Common Pitfalls to Avoid

1. Forgetting Context: Your final sentence must mention the real-world situation (seeds, coins, votes, etc.).
2. Calculation Errors: Remember that \(P(X \ge 8)\) is calculated on your calculator as \(1 - P(X \le 7)\).
3. Hypothesis Symbols: Always use \(p\) for the population proportion. Do not use \(x\) or \(\mu\).
4. Rounding: Don't round your p-values too early; keep them accurate to compare with the significance level.

Key Takeaway: A hypothesis test isn't about being 100% certain. It's about deciding if a result is "too unlikely" to have happened by chance under the current rules. Use your calculator to find those probabilities, compare them to the "bar" (significance level), and write a clear conclusion!