In short
Outlier detection identifies observations far from the rest of the data, for example values more than 3 standard deviations from the mean, or more than 1.5 times the interquartile range beyond the quartiles.
Key takeaways
- Outliers can be data errors, unusual but valid transactions or signs of fraud.
- They should be investigated before being excluded.
- Visualizations such as box plots and scatter plots make them easy to spot.
Where it's tested
BAR: Data Analytics for Accounting — study notes, a five-minute revision sheet and practice questions.
Practise: 3 questions
Question 1 of 3 · 0 answered
An auditor computes a z-score for every expense transaction and flags those with an absolute z-score above 3. What is the auditor doing?