Discussion:
Excluding user defined missing values (=conditional NAs) from multiple imputation in SPSS
(too old to reply)
Adam Smith
2020-01-21 16:48:36 UTC
Permalink
I've got a dataset with survey data. Since some items are not applicable to every participant, I assigned a user defined missing value (i.e. 990). If an item was applicable for a participant, but was not filled in, then I assigned a user defined missing value of 999 to it. So when running a multiple imputation, I want all 999-values to be included in the imputation (and therefore all 999-values should be replaced by calculated values), but 990-values shouldn't be replaced at all! How could this be done? I am wondering that I've found no answer for this problem although one might assume that this procedure is often requested.
I posted it on stackexchange also.
Adam Smith
2020-01-21 16:49:30 UTC
Permalink
Forgot the link to stackexchange: https://stats.stackexchange.com/questions/445690/excluding-user-defined-missing-values-conditional-nas-from-multiple-imputatio
Bruce Weaver
2020-01-21 22:43:31 UTC
Permalink
Post by Adam Smith
I've got a dataset with survey data. Since some items are not applicable to every participant, I assigned a user defined missing value (i.e. 990). If an item was applicable for a participant, but was not filled in, then I assigned a user defined missing value of 999 to it. So when running a multiple imputation, I want all 999-values to be included in the imputation (and therefore all 999-values should be replaced by calculated values), but 990-values shouldn't be replaced at all! How could this be done? I am wondering that I've found no answer for this problem although one might assume that this procedure is often requested.
I posted it on stackexchange also.
This sounds pretty similar (if not identical) to the problem being discussed here:

http://spssx-discussion.1045642.n5.nabble.com/Multiple-imputation-adding-constraints-to-nominal-data-td5738789.html

That forum is the SPSSX-L mailing list. Nowadays, it gets a lot more traffic than this old usenet group does. The actual mailing list is administered by U of Georgia, but the link I gave you is for the Nabble archive. The main page is http://spssx-discussion.1045642.n5.nabble.com/. Click the <more options> link at the top for info on how to join the mailing list. HTH.
Adam Smith
2020-01-22 17:34:39 UTC
Permalink
Thank you. Though it looks a lot like I'm gonna have to do it in R …
Loading...