Which imputation strategy is most appropriate for a numerical feature that exhibits a highly skewed distribution?

Median imputation is robust to outliers and skewed distributions because the median is a measure of central tendency that is not pulled by extreme values. Mean imputation is sensitive to outliers, which can significantly bias the imputed value in skewed data. Mode imputation is typically reserved for categorical data, and zero imputation introduces artificial bias unless zero is a meaningful value in the domain.