Chi-squared tests

Introduction to Chi-Squared (\(\chi^2\)) Tests

Welcome to one of the most practical chapters in Statistics! Have you ever wondered if your "luck" at rolling a die is actually just a weighted die, or if the music people listen to is truly linked to their age? Chi-squared (\(\chi^2\)) tests are the tools mathematicians use to answer these "Expectation vs. Reality" questions.

In this chapter, we will learn how to measure the difference between what we actually see (Observed) and what we expect to see (Expected). If the difference is huge, something interesting is happening!

1. The Core Formula: Measuring the Gap

Every Chi-squared test uses the same basic engine to calculate a test statistic. Don't worry if it looks intimidating; it's just a way to see how "wrong" our expectations were.

\(\chi^2_{calc} = \sum \frac{(O - E)^2}{E}\)

Where:
\(O\) = Observed frequency (the real data you collected).
\(E\) = Expected frequency (what should happen if your theory is true).

An Everyday Analogy:
Imagine you expect to get 10 texts a day (\(E=10\)). One day you get 15 (\(O=15\)). The "gap" is 5. We square that gap (\(5^2 = 25\)) so that negative differences don't cancel out positive ones, and then we divide by the original expectation to see how significant that gap is relative to the total.

Quick Review:
- Large \(\chi^2\) value = Big difference between data and theory.
- Small \(\chi^2\) value = Data fits the theory well.

2. Chi-Squared Test for Association (Contingency Tables)

This test is used when you have categorical data and want to know if two factors are independent. For example: Is "Success in Math" independent of "Whether you eat breakfast"?

Hypotheses

\(H_0\): There is no association between the factors (they are independent).
\(H_1\): There is an association between the factors (they are dependent).

Calculating Expected Frequencies (\(E\))

For each cell in your table, calculate:
\(E = \frac{\text{Row Total} \times \text{Column Total}}{\text{Grand Total}}\)

Degrees of Freedom (\(df\))

The "Degrees of Freedom" tells us how many cells in our table are "free to vary" before the totals lock the rest in place.
Formula: \(df = (r - 1)(c - 1)\)
(Where \(r\) is the number of rows and \(c\) is the number of columns)

Common Mistake to Avoid:
When calculating \(df\), do not count the "Total" row or column! Only count the categories themselves.

Key Takeaway:

If your calculated \(\chi^2\) is greater than the critical value from the table, you reject \(H_0\) and conclude there is an association.

3. Chi-Squared Goodness of Fit Test

This test checks if a specific mathematical model (like a Uniform, Binomial, or Poisson distribution) actually fits your real-world data.

Hypotheses

\(H_0\): The data fits the model (e.g., the Poisson distribution is a suitable model).
\(H_1\): The data does not fit the model.

The Golden Rule: Small Expected Frequencies

Chi-squared tests aren't very reliable if the Expected frequency (\(E\)) is too small.
The Rule: If any \(E < 5\), you must combine that cell with an adjacent cell (and do the same for the observed values).
Don't worry if this seems tricky; just remember: "If it's under five, combine to survive!"

Calculating Degrees of Freedom (\(df\)) for Models

This is slightly different from contingency tables:
\(df = (\text{Number of cells after combining}) - 1 - (\text{Number of parameters estimated from the data})\)

Parameters estimated:
- Uniform: Usually 0 parameters estimated.
- Poisson: 1 parameter (\(\lambda\)) if you calculate the mean from the data.
- Binomial: 1 parameter (\(p\)) if you calculate the probability from the data.

Did you know?

The Chi-squared distribution is always positive and skewed to the right. As the degrees of freedom increase, the shape becomes more like a normal distribution curve!

4. Interpreting the Results

Once you have your \(\chi^2_{calc}\), you have two ways to make a decision:

Method A: Using Critical Value Tables

1. Choose a significance level (usually 5%).
2. Find the Critical Value in the formula booklet using your \(df\).
3. If \(\chi^2_{calc} > \text{Critical Value}\), the result is significant. Reject \(H_0\).

Method B: Using p-values (Software Output)

If you use a calculator or computer, it might give you a p-value.
- If \(p < \text{Significance Level}\): Reject \(H_0\).
- If \(p > \text{Significance Level}\): Fail to reject \(H_0\).

Memory Aid:
"If the p is low, the \(H_0\) must go. If the p is high, the \(H_0\) can fly (stay)."

5. Final Summary Checklist

When tackling a Chi-squared exam question, follow these steps:
1. State Hypotheses clearly (\(H_0\) is always the "no change" or "independence" side).
2. Calculate Expected Values (\(E\)) for every category.
3. Check the \(E \ge 5\) rule. Combine cells if necessary!
4. Find the Test Statistic using the formula \(\sum \frac{(O - E)^2}{E}\).
5. Determine \(df\)** based on the test type.
6. Compare with the critical value and conclude in context (e.g., "There is sufficient evidence to suggest that...").

Quick Review Box:
- Contingency Table \(df\): \((r-1)(c-1)\).
- Expected values: Must be \(\ge 5\).
- Conclusion: Always relate it back to the original words in the question!