How to replace missing data points with Multiple Imputation Data Sets

[mohsin99]

Hello all,
I have successfully imputed 20 data sets using SPSS 20 Multiple Imputation; and have got pooled estimates of mean, SD, SE, variance etc. However, I am still unable identify that how to replace empty/missing data points in original data set? Do I need to take average of each imputed data point (which originally was missing in original data set) from all imputed data sets to replace it in original data set ?
I will thankful if anyone can help me out.
Regards
Mohsin

[Bruce Weaver]

There is no replacement of missing data points in the original data set.  MI works exactly the way you've described:  I.e., the analysis (whatever it is) is done on each of the imputed data sets, and the results are pooled (using the method described by Rubin).  

mohsin99 wrote

Hello all,
I have successfully imputed 20 data sets using SPSS 20 Multiple Imputation; and have got pooled estimates of mean, SD, SE, variance etc. However, I am still unable identify that how to replace empty/missing data points in original data set? Do I need to take average of each imputed data point (which originally was missing in original data set) from all imputed data sets to replace it in original data set ?
I will thankful if anyone can help me out.
Regards
Mohsin

[Jon K Peck]

As Bruce said, the nature of multiple imputation is that there is no single replacement for a missing data point.  That is why the dataset is replicated several times (20 is really a lot!).  If you want a single imputed value, you can use the MVA procedure, but that collapses the distribution of missing values to a single point, so you lose the uncertainty that the multiple imputation process retains.


Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: [phone]




From:        mohsin99 <[hidden email]>
To:        [hidden email],
Date:        05/14/2013 05:52 PM
Subject:        [SPSSX-L] How to replace missing data points with Multiple              Imputation Data              Sets
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Hello all,
I have successfully imputed 20 data sets using SPSS 20 Multiple Imputation;
and have got pooled estimates of mean, SD, SE, variance etc. However, I am
still unable identify that how to replace empty/missing data points in
original data set? Do I need to take average of each imputed data point
(which originally was missing in original data set) from all imputed data
sets to replace it in original data set ?
I will thankful if anyone can help me out.
Regards
Mohsin



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/How-to-replace-missing-data-points-with-Multiple-Imputation-Data-Sets-tp5720243.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[mohsin99]

Thanks Bruce and Jhon. Now, I got the whole idea.
Thanks again.