3  The Outlier Is Part of the Data

data = [1, 1, 2, 1, 2, 1, 1, 3, 1, 87]

What if 87 was generated by exactly the same process as the rest?

Four terms this book keeps separate:

• outlier — likely from a different process (error, contamination)
• rare event — low-probability; says nothing about tail shape
• skew — asymmetry; says nothing about tail weight
• fat tail — a property of the generating process itself

3.1 Summary

Opens with a deliberately uncomfortable dataset:

data = [1, 1, 2, 1, 2, 1, 1, 3, 1, 87]

The instinct is to call 87 an outlier and move on. The chapter blocks that instinct: what if 87 was generated by exactly the same process as the rest? It introduces four terms the entire book will keep separating, because collapsing them is the single most common statistical error the book targets:

def classify_the_question(term):
    definitions = {
        "outlier": "a value likely from a DIFFERENT process (error, contamination)",
        "rare event": "a LOW-PROBABILITY value — says nothing about tail shape",
        "skew": "ASYMMETRY of a distribution — says nothing about tail weight",
        "fat tail": "a property of the GENERATING PROCESS — extremes are expected, just infrequent",
    }
    return definitions[term]

A fair coin landing heads 20 times running is rare and thin-tailed. A skewed exam-score distribution can still have a tail that vanishes fast. The chapter’s closing image: the 87 isn’t noise, it’s information about the shape of the process — and the rest of the book is about learning to read that shape.