Walk me through how you would handle a dataset with 30% missing values in a key demographic variable at MIT.
