Outliers and context

Module 15 quiz

Practice questions: choose an answer and check it to see the explanation. Nothing here is timed or graded.

1. 1. A $48,000 sale is far above the usual range. What should a cleaning pipeline do?

   Delete itCap it at the 99th percentileFlag it for review and keep itReplace it with the median
   Check
2. 2. Refund rows have negative amounts. Are they errors?

   Yes, amounts must be positiveNot necessarily; negative amounts are valid refunds under the rulesOnly if larger than the saleYes, unless rounded
   Check
3. 3. A rule flags amounts "greater than 5000.00" but the code uses >=. What changes?

   NothingAn amount of exactly 5000.00 is wrongly flaggedRefunds are flaggedLarge amounts are no longer flagged
   Check
4. 4. A sale of 48,000 appears in a shop where most sales are under 100. What should cleaning do?

   Delete itFlag it for review and keep itCap it at 100Replace it with the median
   Check
5. 5. Why are negative amounts not automatically errors in a transactions table?

   They always areRefunds and reversals are legitimate negative valuesNegatives are rounding noiseThey should be made positive
   Check
6. 6. A pipeline removes values more than 3 standard deviations from the mean before reporting. What is the risk?

   NoneIt silently changes the population and can remove real, important casesIt is too slowIt keeps too many rows
   Check
7. 7. What is the difference between flagging and filtering an outlier?

   NoneFlagging keeps the row with a marker; filtering removes it from the dataFiltering is reversibleFlagging deletes it
   Check
8. 8. Sensor readings jump to 9999. The device manual says 9999 means "no reading". What should cleaning do?

   Keep 9999 as a valueConvert 9999 to missing with a documented reasonDelete the deviceAverage neighbours
   Check
9. 9. Values 10, 11, 12, 13, 500: which statistic is barely affected by the 500?

   MeanMedianSumMaximum
   Check