Outliers & Scaling

• Outliers vs anomalies


  Outliers are valid pieces of data which lies outside the other data, anomalies are pieces of data with a provable error in the data.
• Outliers formulae


  Working with medians and IQR:

  LQ - 1.5 x IQR, UQ + 1.5 x IQR

  Working with means and sd:

  x̄ - 3sx, x̄ + 3sx
• Scaling - a technique used to change the original data and recalculate the average and measure of spread.
• Scaling with addition / subtraction


  • the average changes
  • the measure of spread does not change
• Scaling with multiplication / division

  • the average changes
  • the measure of spread changes