Course contentsShow
← Data Science
Lesson 200 of 2,1456. The Normal Distribution in DepthFree lesson

Comparing Values Across Different Distributions

Using z-scores to compare observations from different normal distributions on a common scale.

Comparing Values Across Different Distributions

What you'll learn: How to use z-scores to fairly compare observations from completely different datasets or measurements.

The Problem with Direct Comparisons

Imagine you scored 85 on a math test and 78 on a history test. Which performance was better? You can't tell just by looking at the raw scores—what if the math test was easier and everyone scored high, while the history test was harder?

This is where z-scores become incredibly powerful for comparison.

Why Z-Scores Enable Fair Comparisons

A z-score tells you how many standard deviations above or below the mean an observation falls. Since it's measuring relative position within a distribution, you can compare z-scores even when the original measurements are in completely different units or scales.

Example: Suppose the math test had μ = 80, σ = 5, while the history test had μ = 70, σ = 4.

  • Math z-score: z = (85 - 80) / 5 = +1.0
  • History z-score: z = (78 - 70) / 4 = +2.0

Even though your raw history score was lower, your relative performance was better—you were 2 standard deviations above average in history but only 1 standard deviation above in math.

Real-World Applications

  • Comparing students across different grading scales (percentages vs. 4.0 scale)
  • Evaluating athletes in different sports (basketball points vs. baseball batting average)
  • Financial analysis (comparing stocks with vastly different price ranges)
  • Medical measurements (comparing height, weight, and blood pressure on the same scale)

Z-scores transform everything into a common, standardized scale where you're always comparing "how unusual is this value relative to its group?"

Key Takeaway: Z-scores let you compare apples to oranges by converting both into a universal measure of relative standing within their respective distributions.