Hi Mil, 

It sounds like the overimpute() function is the one you want here, though, it doesn't actually return the overimputations, just plots them. Here's a rough breakdown of each:

overimpute() - for a particular variable, use the results of the imputation model to generate imputations of each *observed* value, then plot the results. You don't get the actual overimputations here.

overimp - passing a matrix of row/columns to this argument will tell Amelia to treat these values as missing and impute them as part of the imputation. 

overvalues - this is part of the Amelia output where we save the original observed values being overimputed through 'overimp.'

Cheers,
Matt



On Fri, Nov 14, 2014 at 9:41 PM, Mil Jacobs <mil@miljacobs.com.au> wrote:
Hi Matt,

Basically what I want to do is a post estimation type command.  I want to first impute all the missing values and then use the model to calculate theoretical 'imputed' values for the complete cases.  However, I only want m imputed values for each complete case - not hundreds or thousands of values.

What does overimp do exactly?  How is it different from the overimpute command? There is also something in the documentation called overvalues.  Could you clarify each of these? 

So far I have found Amelia very fast and easy to use - looking forward to the updated version :)

Rgds,

On Sat, Nov 15, 2014 at 3:53 AM, Matt Blackwell <mblackwell@gov.harvard.edu> wrote:
Hi Mil, 

I'm not sure I know what you want to do. Do you want to do a leave-one-out approach and overimpute each individual cell at a time? Or do you want to overimpute all the data at the same time. The latter isn't possible since there wouldn't be any data to impute. 

A overimputation one-at-a-time strategy could be done, but you'd have to rerun Amelia each time with overimp set differently each time. Note that there is a bug in amelia on CRAN right now that makes overimp sometimes not work. This should be fixed in the next version of Amelia, to be sent to CRAN in the next few days. 

Hope that helps!

Cheers,
Matt

On Thu Nov 13 2014 at 10:13:47 PM Mil Jacobs <mil@miljacobs.com.au> wrote:
I would like to calculate imputed values for the cases that are complete in my data set but I am not sure how to do this.  

I do not want thousands of overimputations, just one imputed value for each complete case in each imputed data set (in the same way as for the missing cases).  How do I do this?  I have tried using 'overimp' with the 'amelia' command but it is not working.



 --
Amelia mailing list served by HUIT
[Un]Subscribe/View Archive: http://lists.gking.harvard.edu/?info=amelia
More info about Amelia: http://gking.harvard.edu/amelia
Amelia mailing list
Amelia@lists.gking.harvard.edu

To unsubscribe from this list or get other information:

https://lists.gking.harvard.edu/mailman/listinfo/amelia