Speaking Stata: Replacing missing values: The easiest problems

Abstract

Missing values are common in real datasets, and what to do about them is a large and challenging question. This column focuses on the easiest problems in which a researcher is clear, or at least highly confident, about what missing values should be instead, implying a deterministic replacement. The main tricks are copying values from observation to observation and using the ipolate command. Both may often be extended simply to panel or longitudinal datasets or to other datasets with a group structure, such as data on individuals within families or households. This column includes how to satisfy constraints that interpolation is confined to filling gaps between values known to be equal or to observations moderately close to a known value in time or in some other sequence or position variable.