It has been repeatedly emphasized that the imputed values should not be

treated as the real or actual values for the nonrespondents. However, if a

singly imputed data set is analyzed using any standard software, then they

are treated as real values, therefore, the stated standard errors will understate

the uncertainty. The fact that imputations are guesses (or plausible values)

created from the observed data should be reflected in the standard error cal-

culations. Multiple imputation is a mechanism for adding this additional

uncertainty.