Exploratory Data Analysis

Deletion (dropping rows or columns with missing values) and imputation (filling in estimated values). Deletion is simple but can discard useful data; imputation preserves sample size but can introduce bias.

When missingness is small relative to the dataset and the data are missing completely at random (MCAR), so dropping those rows does not bias the remaining sample.

Replacing missing values with a column summary statistic: mean for roughly symmetric numeric data, median for skewed numeric data or data with outliers, and mode for categorical data.

It artificially reduces variance and weakens correlations because every missing value collapses to the same central value, creating a spike at the mean.

Using a predictive model (e.g., KNN, regression, or MICE) trained on the other features to estimate missing values, capturing relationships between variables rather than using a single summary statistic.