Course contentsShow
← Data Science
Lesson 71 of 2,1452. Descriptive Statistics FoundationsFree lesson

Z-Score Method for Outlier Detection

Apply z-scores to detect outliers by measuring standard deviations from the mean in normally distributed data.

Z-Score Method for Outlier Detection

You'll learn: How to use z-scores to mathematically identify data points that are unusually far from your dataset's mean.

The Core Idea

Remember z-scores? They tell you how many standard deviations a value sits from the mean. The z-score method uses this to flag outliers: if a data point is too many standard deviations away from the mean, it's probably an outlier.

The standard rule of thumb: any z-score beyond ±3 is typically considered an outlier. Some analysts use ±2.5 or ±2 for stricter detection, depending on context.

How It Works

  1. Calculate the mean and standard deviation of your dataset
  2. For each data point, compute its z-score: how many standard deviations it is from the mean
  3. If the absolute z-score exceeds your threshold (usually 3), flag it as an outlier

Analogy Time

Imagine a classroom where the average test score is 75 with a standard deviation of 10. A student scoring 105 is exactly 3 standard deviations above average (z-score = +3). That's extremely rare in a normal distribution—unusual enough to investigate. Did they study exceptionally hard, or was there an error?

When to Use This Method

The z-score approach works best when your data is normally distributed (bell-shaped). It assumes the mean and standard deviation are meaningful measures, which isn't true for heavily skewed data. For non-normal distributions, the IQR method (box plots) is often more appropriate.

Why It Matters

This method gives you a quantitative, objective way to identify outliers rather than just "eyeballing" visualizations. It's reproducible and transparent—anyone can verify your threshold and recalculate.

Key Takeaway: The z-score method detects outliers by flagging values more than 3 standard deviations from the mean, providing a statistical foundation for outlier identification in normally distributed data.