If somebody had some experience of how to deal with the following missing data 
situation: 

 We have a survey that use the systematic sampling and there are total of 258 
records. 
There are 98 cases that have at least some missing data on the 3 outcome 
measures. 
There are 33 cases that have more than 10% of the items of the outcome 
variables missing. 
Here comes the analysis, we have different ideas and don't know which one is 
correct: 
One idea is to delete those 33 cases from the inferential analyses... 
Then, the next step is to "replace" the random items that are missing on the 
other 65 cases. 
The question is; 1. should we do the imputation for this analysis? Or just 
delete the missing records?
2. if we do the imputation, should we do just once or should we do the ITERATE? 
We are argued that the entire idea of imputation of missing values is to 
ITERATE through replacement. Thus, we would replace once, do regression, 
replace again, do regression, etc., and then average over the results. If you 
only replace missing values once, your confidence intervals are incorrect. If 
you are going to replace values by the mean/median of local results, you have 
to somehow account for this in subsequent analysis. you can't just then analyze 
the "replaced" data as if they were the true results.  Because there is no 
reason why anyone's record should be related to anyone else's.
Any ideas?


 Julia Zhu, MS 
ASSOCIATE SERVICE FELLOW 
Statistics 
CDC/OD/OWCD/Science Office 
(Office) 404-498-2382 
(Fax) 404-498-6365 
MS E-94
(Email) [email protected] 



 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://lists.utsouthwestern.edu/pipermail/impute/attachments/20090225/4c4b12bd/attachment.htm
From drj10 <@t> psu.edu  Thu Feb 26 19:44:38 2009
From: drj10 <@t> psu.edu (David R Johnson)
Date: Thu Feb 26 19:44:43 2009
Subject: [Impute] The new Multiple Imputation procedure  in SPSS 17
Message-ID: <[email protected]>

Has anyone used the new multiple imputation module in SPSS?  We have 
been trying it out and it seems to be pretty fast, but the documentation 
is very sparse.  We would be interested in hearing other folks 
experiences with this module. 

David J.

-- 
David R. Johnson
Professor of Sociology, Human Development and Family Studies, and Demography
Department of Sociology
713 Oswald Tower
The Pennsylvania State University
University Park, PA 16802
tel:  814-865-9564

Reply via email to