Variance and standard deviation for conditional and marginal distributions

Welcome to the World of Variability!

In our journey through Multivariate Random Variables, we’ve already learned how to find the "average" (expected value) of one variable when we know something about another. But as any actuary will tell you, the average is only half the story. To truly manage risk, we need to know how much things "spread out" around that average. That’s where Variance and Standard Deviation come in.

In this chapter, we are going to look at these concepts through two lenses: 1. Marginal: Looking at one variable while ignoring the other. 2. Conditional: Looking at how the spread of one variable changes when we have specific information about the other.

Don't worry if this seems a bit abstract right now. We'll break it down step-by-step with simple analogies!

1. Marginal Variance and Standard Deviation

When we talk about Marginal Variance, we are "zooming out." Even though we have a joint distribution (X and Y together), we choose to focus only on one variable and ignore the other completely.

How to calculate it:

To find the variance of \( X \), denoted as \( Var(X) \), you first need the Marginal Distribution of \( X \). If you have a joint PDF or PMF, you "sum out" or "integrate out" the variable \( Y \).

Once you have the marginal distribution, the formula is the same one you learned in univariate probability:
\( Var(X) = E[X^2] - (E[X])^2 \)

The Standard Deviation is simply the square root of the variance:
\( \sigma_X = \sqrt{Var(X)} \)

Quick Review: Remember that \( E[X^2] \) for a continuous variable is \( \int x^2 f_X(x) dx \) and for a discrete variable is \( \sum x^2 p_X(x) \).

Key Takeaway: Marginal variance treats the variable as if its partner doesn't exist. It measures the total "risk" or "spread" of that single variable across the entire population.

2. Conditional Variance

This is where things get interesting! Conditional Variance, denoted as \( Var(Y|X=x) \), measures the spread of \( Y \) given that we already know the exact value of \( X \).

Think of it this way:

Imagine you are looking at the height of adults (\( Y \)). The Marginal Variance is the spread of heights for everyone. But if I tell you the person is 12 years old (\( X=12 \)), the Conditional Variance is the spread of heights only among 12-year-olds. Usually, the spread among 12-year-olds is much smaller than the spread of the whole world!

The Formula:

The logic remains the same as standard variance, but we use conditional expectations:
\( Var(Y|X=x) = E[Y^2|X=x] - (E[Y|X=x])^2 \)

Step-by-Step Process:

1. Find the conditional PDF/PMF: \( f_{Y|X}(y|x) = \frac{f(x,y)}{f_X(x)} \).
2. Calculate the conditional mean: \( E[Y|X=x] \).
3. Calculate the conditional second moment: \( E[Y^2|X=x] \).
4. Plug them into the formula above.

Common Mistake to Avoid: Remember that \( Var(Y|X=x) \) is usually a function of x, not just a single number (unless \( X \) and \( Y \) are independent). If you plug in a specific number for \( x \), you get a number. If you leave it as the variable \( X \), you get a new random variable!

3. The Law of Total Variance (The "EVVE" Rule)

This is arguably the most important formula in this chapter for Exam P. It connects marginal variance and conditional variance. It allows us to find the total variance of \( Y \) by breaking it into two pieces based on \( X \).

The Formula:

\( Var(Y) = E[Var(Y|X)] + Var(E[Y|X]) \)

Memory Aid: "EVVE"
Expected Value of the Variance + Variance of the Expected Value.

What does this actually mean?

Let's use an analogy. Suppose you want to find the variance in temperature across the entire United States today (\( Var(Y) \)).

1. \( E[Var(Y|X)] \) (The "Within" group variance): This is the average of the variances within each city. On average, how much does the temperature fluctuate within a single city?
2. \( Var(E[Y|X]) \) (The "Between" group variance): This is the variance of the average temperatures of the cities. How much does the average temperature of Miami differ from the average temperature of Anchorage?

Did you know? This is often called the Variance Decomposition Formula because it decomposes total risk into "unexplained randomness" and "predictable variation."

Key Takeaway: If you need \( Var(Y) \) and the problem gives you conditional information, don't panic! Just find the pieces for the EVVE formula.

4. Important Properties and Relationships

Here are a few "pro-tips" for the exam:

1. Independence: If \( X \) and \( Y \) are independent, then knowing \( X \) tells you nothing about \( Y \). Therefore:
\( E[Y|X] = E[Y] \)
\( Var(Y|X) = Var(Y) \)

2. Standard Deviation is NOT Additive:
Never forget: \( \sigma_{Y|X} = \sqrt{Var(Y|X)} \). You cannot simply add standard deviations together. Always do your math with Variance first, then take the square root at the very end.

3. The "Inner" Variable: In the expression \( Var(E[Y|X]) \), the term \( E[Y|X] \) is a random variable that is a function of \( X \). When you take the variance of it, you are treating it just like any other function \( g(X) \).

Quick Summary Checklist

- [ ] Marginal Variance: Use the marginal distribution; ignore the other variable.
- [ ] Conditional Variance: The spread of one variable when the other is fixed (\( x \)).
- [ ] EVVE Rule: \( Var(Y) = E[Var(Y|X)] + Var(E[Y|X]) \).
- [ ] Function vs. Value: \( Var(Y|X=x) \) is a number; \( Var(Y|X) \) is a random variable.
- [ ] Final Step: Standard deviation is always the square root of the variance.

Don't worry if the Law of Total Variance feels like a mouthful at first. Practice a few problems where you have to calculate the "Within" and "Between" components separately, and you'll start to see the pattern! You've got this!