Inference using normal and t-distributions - Mathematics - Further (9231) - Cambridge International AS Level

Welcome to Statistical Inference!

Ever wondered how scientists can claim a new medicine works by only testing it on a few hundred people? Or how pollsters predict election results before the final vote is counted? That is the power of Statistical Inference!

In this chapter, we are moving from simply describing data to making "educated guesses" about a whole population based on a small sample. Don't worry if it sounds a bit like magic—it’s actually all about logic and two very special distributions: the Normal (Z) distribution and the t-distribution.

Prerequisite Check: Before we dive in, remember that the mean ($\mu$) is the average, and the variance ($\sigma^2$) tells us how spread out the data is. If you've seen the "Bell Curve" before, you're already halfway there!

1. The Big Picture: Point vs. Interval Estimates

If I ask you how much a giant pumpkin weighs, you might give me one number (e.g., "50kg"). That’s a point estimate. But if you want to be more reliable, you might say, "It's between 45kg and 55kg." That’s an interval estimate.

In this course, we use the sample mean ($\bar{x}$) as our point estimate for the population mean ($\mu$). Because samples aren't perfect, we build a Confidence Interval (CI) around that mean to say how sure we are.

2. The "Z" vs. "t" Decision: Which one do I use?

This is the most important choice you will make in every exam question. Using the wrong table will lead to the wrong answer! Here is the simple rule of thumb:

Use the Normal (Z) Distribution if:

1. You know the population variance ($\sigma^2$).
2. You don't know the variance, but your sample size is large ($n \geq 30$). In this case, we use the Central Limit Theorem and use the sample variance $s^2$ as an estimate.

Use the t-Distribution if:

1. You don't know the population variance ($\sigma^2$).
2. Your sample size is small ($n < 30$).
3. Crucial Requirement: The population must be normally distributed for the t-test to be valid!

Quick Review Box:
- Known $\sigma^2$ $\rightarrow$ Z
- Unknown $\sigma^2$ + Large $n$ $\rightarrow$ Z
- Unknown $\sigma^2$ + Small $n$ $\rightarrow$ t

3. Confidence Intervals for a Single Mean

A confidence interval gives us a range where we believe the true population mean lies. The formula looks like this:

$\bar{x} \pm (\text{Critical Value}) \times \frac{s}{\sqrt{n}}$

Step-by-Step Process:
1. Find $\bar{x}$: The average of your sample.
2. Find $s^2$: The unbiased estimate of the population variance (if not given). Use the formula $s^2 = \frac{n}{n-1} \times (\text{sample variance})$.
3. Find the Critical Value:
- For Z, look up the percentage in your normal tables (e.g., 1.96 for 95%).
- For t, you need Degrees of Freedom ($\nu = n - 1$). Look at the t-table under your chosen significance level.
4. Calculate the Margin of Error: Multiply the critical value by the standard error ($\frac{s}{\sqrt{n}}$).
5. State the Interval: (Lower Bound, Upper Bound).

Example: If you measure 10 chocolate bars and find a mean weight of 50g with $s = 2$, your "degrees of freedom" for a t-test would be $10 - 1 = 9$.

Key Takeaway: A 99% Confidence Interval will be wider than a 95% interval because you are being "more sure," which requires a bigger safety net!

4. Hypothesis Testing: The 5-Step Logic

Hypothesis testing is like a court trial. We assume the "Null Hypothesis" ($H_0$) is true (Innocent) until we have enough evidence to prove the "Alternative Hypothesis" ($H_1$) (Guilty).

The Steps:

1. State the Hypotheses:
- $H_0: \mu = \text{value}$
- $H_1: \mu \neq, <, \text{ or } > \text{value}$

2. Calculate the Test Statistic:
- Using Z: $z = \frac{\bar{x} - \mu}{\sigma / \sqrt{n}}$
- Using t: $t = \frac{\bar{x} - \mu}{s / \sqrt{n}}$

3. Determine the Critical Value or p-value:
Use your tables based on the significance level (usually 5% or 1%).

4. Compare:
If your calculated value is further away from zero than the critical value, it’s "unlikely" to have happened by chance.

5. Conclude in Context:
Don't just say "Reject $H_0$." Say, "There is significant evidence to suggest the mean weight of chocolate bars has decreased."

5. Comparing Two Means (Independent Samples)

Sometimes we want to know if two groups are different—for example, "Do boys score higher than girls on this test?"

For this, we look at the difference in means: $(\bar{x}_1 - \bar{x}_2)$.

When using the t-distribution for two groups:

In Further Math 9231, we usually assume the two populations have the same variance. We combine (pool) their variances to get a Pooled Estimate ($s_p^2$).

The Pooled Variance Formula:
$s_p^2 = \frac{(n_1 - 1)s_1^2 + (n_2 - 1)s_2^2}{n_1 + n_2 - 2}$

Degrees of Freedom: For two samples, $\nu = n_1 + n_2 - 2$.

Common Mistake: Students often forget to add the two sample sizes and subtract 2. Remember: Two samples, two "lost" degrees of freedom!

6. Matched Pairs (The "Before and After" Test)

Imagine testing a diet. You weigh 10 people before and after. These aren't independent groups; they are the same people! This is a Matched Pairs t-test.

The Trick:
1. Calculate the difference ($d$) for each person (After minus Before).
2. Treat these differences as your new single sample.
3. Test if the mean of these differences ($\mu_d$) is zero.
4. Use $\nu = n - 1$, where $n$ is the number of pairs.

Did you know? Using a matched pairs test is much more powerful than an independent test because it ignores the differences between people and focuses only on the change within each person.

7. Summary Checklist for Success

- Check your $n$: Is it small or large?
- Check your $\sigma^2$: Is it known or estimated?
- State your assumptions: If using a t-test, always write: "Assume the population is normally distributed."
- Read the tail: Is it a one-tailed test (e.g., "increase") or two-tailed (e.g., "change")?
- Context is King: Always relate your final answer back to the pumpkins, chocolate bars, or test scores mentioned in the question!

Don't worry if this seems tricky at first! Statistical inference is like learning a new language. Once you get the "grammar" (the 5 steps) down, the "vocabulary" (the formulas) will start to make perfect sense.

* The content provided by thinka is generated by AI and may not always be accurate or up-to-date. Please use it as a supplementary resource and verify with official materials.