0
$\begingroup$

After multiple imputation, the missing values are provided with a imputed valid data which appear as a separate entity. So I corrected this by using the recode to same variables. Instead of using fixed values. For example, 0=0, I used a range because multiple imputation uses a range and thus recoded them as 0 through 1 to 1 and so on. I got the answer and all the imputed data was grouped into the intended categories of the variables.

I would like to know if it is best practice to do that. Is there any reference that using range for multiple imputation is good.

|cite|improve this question
$\endgroup$
  • $\begingroup$ Can you provide a little more information about the variable that has missing data? In particular, is it an ordered categorical variable? And when you say that you got the answer and all the imputed data were grouped into the intended categories, do you mean that this is what SPSS gave you, or is this something you did ex post facto? $\endgroup$ – Matthew Sep 5 '16 at 15:22
  • $\begingroup$ @Matthew I have a variable called frequency of toothbrushing which I recoded as 1 (twice or more) and (2) once or less. It had 30 missing values (out of total of 667). After multiple imputation, the 30 missing values were replaced as valid values (1 or2) but then the problem is that they are being represented as separate entity and they are not grouping into the original levels of 1 and 2. Instead they lie separately. For example, since I labelled 1 and 2, I get twice or more = 300 once or less = 337 and the imputed values lie separately as 1 = so and so and 2 = so on. $\endgroup$ – James Rufus Sep 6 '16 at 0:15
  • $\begingroup$ Do you mean they just have different labels but the underlying values are 1 and 2? (I don't use SPSS so I'm not familiar with how it out-puts data after MI) $\endgroup$ – Matthew Sep 6 '16 at 7:19
1
$\begingroup$

Multiple Imputation already adds randomness in the different imputations, so I don't seen the point of overriding that.

|cite|improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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