ABSTRACT
An imputation based approach involves replacing the set of missing values
with a plausible set of values. This plausible set is created or estimated based
on available information. This task should not be construed as creating “ac-
tual” or “real” values for nonrespondents. For a layman, this idea conjures
an image of a statistician making up the data. This is true only if one were
to analyze the data as if the imputed data are real values. Thus, any sin-
gle imputation analysis that does not account for the imputation process in
the ensuing statistical inferences may inflate the precision and declare effects
significant when they are not, and thus invalid.