Linear Transformations of Random Variables
What you'll learn: How shifting and scaling a random variable affects its probability distribution, expected value, and variance.
The Core Idea
Suppose you have a random variable X (like test scores out of 50) and you want to transform it into a new random variable Y using the formula:
Y = aX + b
where a and b are constants. For example, converting scores to a 100-point scale would use a = 2 and b = 0. This is called a linear transformation.
The beauty is we can predict Y's distribution, mean, and variance directly from X's properties—without starting from scratch.
How It Works
Distribution of Y
If you know how X is distributed, Y inherits a transformed version:
- For discrete X: the PMF shifts and stretches according to y = ax + b
- For continuous X: the PDF scales by 1/|a| and shifts by b
Expectation of Y
The mean transforms predictably:
E[Y] = E[aX + b] = aE[X] + b
Adding a constant b shifts the mean by b. Multiplying by a scales the mean by a.
Variance of Y
The variance only responds to scaling (constants vanish since they don't add randomness):
Var(Y) = Var(aX + b) = a²Var(X)
Notice the a² term—scaling doubles the spread but quadruples the variance. The constant b disappears entirely because shifting doesn't change spread.
Analogy: Think of X as temperatures in Celsius. Converting to Fahrenheit (Y = 1.8X + 32) shifts the mean by 32 and scales the variance by 1.8² = 3.24.
Key Takeaway: For Y = aX + b, the mean transforms linearly (E[Y] = aE[X] + b), while variance scales quadratically (Var(Y) = a²Var(X)), ignoring additive constants.