CompTIA Data+ DA0-002 (V2) Practice Question

Median imputation is generally preferred for a right-skewed numeric variable because the median is resistant to extreme values in the long tail. Replacing NULLs with the mean would pull the imputed values toward the large outliers and artificially raise the typical spending level. Inserting zeros would severely distort both the center and the spread of the data and understate revenue. Forward-fill (last observation carried forward) is designed for ordered time-series data and makes the unrealistic assumption that a customer's previous purchase amount persists unchanged; it does not address the underlying skewness. Therefore, using the column median calculated from the non-NULL observations is the most appropriate single-value imputation in this scenario.