What is Grubbs' Test?
You'll learn: How Grubbs' Test provides a formal statistical hypothesis test for detecting a single outlier in normally distributed data.
The Core Idea
While the Z-Score and IQR methods give you rules of thumb for flagging outliers, Grubbs' Test takes a more rigorous approach. It's a formal hypothesis test specifically designed to detect whether the most extreme value in your dataset is a genuine statistical outlier.
Think of it like this: You've used Z-scores to suspect an outlier, but Grubbs' Test gives you a statistical verdict with a confidence level.
How It Works
Grubbs' Test focuses on one data point at a time—the single most extreme value (either maximum or minimum). It calculates a test statistic based on:
- How far the suspected outlier is from the sample mean
- The sample standard deviation
- The sample size
The test then asks: "If this data truly comes from a normal distribution, how likely is it that we'd see a value this extreme?" If the probability is very low (below your significance level, typically 0.05), you reject the null hypothesis and declare the point an outlier.
Key Assumptions
- Normality required: Your data must be approximately normally distributed (Grubbs' is parametric)
- Single outlier: Designed to detect one outlier at a time
- Independent observations: Data points should be independent
When to Use It
Use Grubbs' Test when you need statistical rigor rather than just a heuristic, and when you can reasonably assume your underlying data follows a normal distribution.
Key Takeaway: Grubbs' Test is a parametric hypothesis test that formally determines whether the most extreme value in normally distributed data is a statistical outlier, providing more rigor than threshold-based methods like Z-score or IQR.