Python for Data Science For Dummies, 2nd Edition

Chapter 16

Detecting Outliers in Data

IN THIS CHAPTER

Understanding what is an outlier

Distinguishing between extreme values and novelties

Using simple statistics for catching outliers

Finding out most tricky outliers by advanced techniques

Errors happen when you least expect, and that’s also true in regard to your data. In addition, data errors are difficult to spot, especially when your dataset contains many variables of different types and scale. Data errors can take a number of forms. For example, the values may be systematically missing on certain variables, erroneous numbers could appear here and there, and the data could include outliers. A red flag has to be raised when:

• Missing values on certain groups of cases or variables imply that some specific cause is generating the error.
• Erroneous values depend on how the application has produced or manipulated the data. For instance, you need to know whether the application has obtained data from a measurement instrument. External conditions and human error can affect the reliability of instruments.
• The case is apparently valid, ...