Describe random sampling and the sampling distributions of statistics commonly used in statistical inference

Welcome to the Bridge of Statistical Inference!

In your CS1 journey so far, you’ve learned about various probability distributions. Now, we are moving into a very exciting area: Statistical Inference. This is where the magic happens! We use a small amount of data (a sample) to make smart guesses about a much larger group (the population).

Think of it like being a chef. You don't need to eat the whole pot of soup to know if it needs more salt; you just need a well-mixed spoonful. In this chapter, we will learn how that "spoonful" (the sample) behaves and how it relates to the whole "pot" (the population). Don't worry if this seems a bit abstract at first—we'll break it down step-by-step!

1. What is Random Sampling?

To make sure our "spoonful of soup" represents the whole pot, every part of the soup must have an equal chance of being chosen. This is the heart of Random Sampling.

In actuarial terms, we say that \(X_1, X_2, ..., X_n\) is a random sample of size \(n\) if:

• Each \(X_i\) comes from the same distribution (they are identically distributed).
• The value of one observation doesn't affect the others (they are independent).

You will often see this abbreviated as i.i.d. (independent and identically distributed). If you see "i.i.d." in an exam question, it’s just code for "this is a perfect random sample."

2. Statistics vs. Parameters

Before we go further, let's clear up two terms that students often mix up:

• Parameter: A fixed (usually unknown) value that describes the whole population. Examples: Population mean (\(\mu\)) or population variance (\(\sigma^2\)).
• Statistic: A value calculated from your sample. Examples: Sample mean (\(\bar{X}\)) or sample variance (\(S^2\)).

Quick Tip: Remember P for Population/Parameter and S for Sample/Statistic!

What is a Sampling Distribution?

Imagine you take a sample of 10 people and calculate their average height. Then you take another sample of 10 people and do the same. You’ll get a slightly different average each time. If you did this thousands of times and plotted all those averages on a graph, the resulting distribution is called the Sampling Distribution of the statistic.

3. The Distribution of the Sample Mean (\(\bar{X}\))

The sample mean is the most common statistic we use. Its behavior depends on the population it comes from.

Case A: Sampling from a Normal Distribution

If your population is already Normal, i.e., \(X \sim N(\mu, \sigma^2)\), then the sample mean is also Normal!

\(\bar{X} \sim N(\mu, \frac{\sigma^2}{n})\)

Key Takeaway: The average of the sample means is the same as the population mean (\(\mu\)), but the spread (variance) is smaller. The larger your sample size (\(n\)), the more certain you are about the result, so the variance decreases.

Case B: The Central Limit Theorem (CLT)

What if the population is NOT Normal? (Maybe it's Exponential or Gamma). This is where the Central Limit Theorem comes to the rescue! It states that if your sample size \(n\) is "large enough" (usually \(n > 30\)), the distribution of the sample mean will be approximately Normal, regardless of the original distribution's shape.

\(\bar{X} \approx N(\mu, \frac{\sigma^2}{n})\)

Analogy: Imagine a crowd of people. Individually, they might be doing weird things (different distributions), but when you look at the "average" behavior of the crowd, it tends to follow a predictable, bell-shaped pattern.

4. The Distribution of the Sample Variance (\(S^2\))

While the mean follows a Normal distribution, the sample variance follows a different path. When we sample from a Normal population, we use the Chi-square (\(\chi^2\)) distribution to describe the variance.

The formula to remember is:

\(\frac{(n-1)S^2}{\sigma^2} \sim \chi^2_{n-1}\)

Where:

• \(n\) is the sample size.
• \(S^2\) is the sample variance.
• \(\sigma^2\) is the population variance.
• \(n-1\) is called the degrees of freedom.

Common Mistake: Students often forget to use \(n-1\) instead of \(n\). In statistics, we "lose" one degree of freedom because we used the sample mean to calculate the variance. Just remember: Variance uses \(n-1\)!

5. The Student’s t-distribution

In the real world, we almost never know the true population variance (\(\sigma^2\)). If we don't know \(\sigma^2\), we have to use the sample variance (\(S^2\)) instead. When we do this, the Normal distribution is no longer perfectly accurate—we use the t-distribution instead.

The statistic is: \(T = \frac{\bar{X} - \mu}{S / \sqrt{n}}\)

This follows a \(t_{n-1}\) distribution (a t-distribution with \(n-1\) degrees of freedom).

Did you know? The t-distribution looks very similar to the Normal distribution but has "fatter tails." This accounts for the extra uncertainty because we are estimating the variance. As \(n\) gets very large, the t-distribution eventually becomes the Normal distribution!

6. Comparing Two Variances: The F-distribution

Sometimes actuaries need to compare the volatility (variance) of two different groups (e.g., claim amounts in two different regions). To do this, we look at the ratio of their sample variances.

If we have two independent samples from Normal populations, the ratio:

\(F = \frac{S_1^2 / \sigma_1^2}{S_2^2 / \sigma_2^2} \sim F_{n_1-1, n_2-1}\)

This follows the F-distribution. It has two different sets of degrees of freedom—one for the top (numerator) and one for the bottom (denominator).

Quick Review Table

Here is a summary to help you keep these straight:

Statistic	Distribution	When to use it?
Sample Mean (\(\bar{X}\))	Normal (\(N\))	Known \(\sigma^2\) or large \(n\) (CLT)
Sample Variance (\(S^2\))	Chi-square (\(\chi^2\))	Inference about \(\sigma^2\) (Normal pop)
Mean with estimated \(S^2\)	t-distribution	Unknown \(\sigma^2\) (Normal pop)
Ratio of two variances	F-distribution	Comparing two Normal populations

Summary and Key Takeaways

• Random Sampling requires observations to be independent and identically distributed (i.i.d.).
• The Sample Mean (\(\bar{X}\)) is Normal if the population is Normal, or approximately Normal if \(n\) is large (thanks to the Central Limit Theorem).
• The Standard Error of the mean is \(\sigma / \sqrt{n}\). Notice it gets smaller as \(n\) increases!
• The Chi-square distribution is the go-to for sample variance.
• Use the t-distribution when \(\sigma^2\) is unknown and you are using \(S^2\) instead.
• Use the F-distribution to compare two variances.

Final Encouragement: You’ve just mastered the language of statistical inference! These distributions are the tools you will use for the rest of the CS1 course to build confidence intervals and perform hypothesis tests. Keep practicing the formulas, and they will soon become second nature!