Question

An analyst is profiling a dataset of 10,000 temperature sensor readings that follow an approximately normal distribution. Before feeding the data into a forecasting model, the analyst must flag readings that are statistically extreme so they can be reviewed. Which technique is most appropriate for identifying these outliers in a normally distributed variable?

Accepted Answer

Calculate a Z-score for each reading and flag any value with an absolute Z-score greater than 3

Answer

Replace every value below the mean with the mean to reduce variability

Answer

Sort the readings alphabetically and remove the top and bottom 100 rows

Answer

Convert all readings to a single categorical bucket labeled 'normal'