1
$\begingroup$

I've completed an EM imputation to replace a small amount of missing data. One of the missing data was within my age variable (age range of 20 - 54).

My question is ... the new score that has been used to replace the missing data within my age variable is an age of '18' ... but my age range was 20-54 originally, and there were never any scores of 18 to begin with. Is this normal for an imputation, or have I done something wrong?

This is using SPSS.

$\endgroup$

1 Answer 1

1
$\begingroup$

I assume in my answer that you are using MVA (Missing Values Analysis). If so, predicted values that fall outside the observed range of the data happens with regression-based imputation strategies like EM (with continuous data specifically). Think of the the imputed values as if they were "predicted values" from a regression equation. In such cases, we regularly see values somewhat out of the range of the observed data.

If you really want to impose constraints on the predicted values, then please consider using the MULTIPLE IMPUTATION command. It features subcommand CONSTRAINTS expressly for such a purpose.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.