Course contentsShow
← Data Science
Lesson 54 of 2,1452. Descriptive Statistics FoundationsFree lesson

When to Use Each Measure

Choose appropriate variability measures based on data distribution and outliers.

When to Use Each Measure

You'll learn how to pick the right variability measure for your data's shape and quirks.

Now that you've met range, variance, standard deviation, IQR, MAD, and coefficient of variation, how do you choose which one to use? The answer depends on your data's distribution and whether outliers are lurking.

Distribution Matters

Think of your data like a crowd of people. If they're evenly spread (symmetric distribution), standard deviation and variance work beautifully—they use every data point and are mathematically elegant. They're the go-to for "well-behaved" data.

But if your crowd has a few giants towering over everyone else (outliers or skewed distribution), standard deviation gets distorted. Those extreme values inflate it, making your spread look bigger than it really is for most of the data.

When Outliers Appear

This is where IQR (Interquartile Range) shines. It ignores the extreme 25% on each end and focuses on the middle 50%. It's resistant to outliers—like wearing noise-canceling headphones in a loud room. Income data, housing prices, and reaction times often have outliers, making IQR the safer choice.

Range is simple but fragile—one extreme value ruins it. Use it only for a quick, rough sense of spread when you know your data is clean.

Special Cases

MAD (Mean Absolute Deviation) is useful when you want interpretability similar to standard deviation but with less sensitivity to outliers.

Coefficient of Variation is your tool when comparing datasets with different units or scales (e.g., comparing variability in temperature vs. income).

Key Takeaway: Use standard deviation for symmetric, outlier-free data; switch to IQR when outliers or skewness appear; choose coefficient of variation for cross-scale comparisons.