Hi Stefanie, 

Amelia has to estimate a large number of parameters: p*(p+3)/2 where p is the number of parameters. Thus, with 121 variables, it must estimate roughly 7500 parameters, which is far greater than the number of observations. You will have to pair down the variables to those that are most crucial to the analysis. 

As for collinearity, this message usually occurs when there are variables in the data frame that are linear combinations of one another. For instance, if you had a variable for male == 1 and a separate variable for female == 1 (and there were no cases that were not one of these two categories). You would have to omit one of these variables. 

Finally, the write.amelia function should write out 5 "csv" files, each with data. Are there 5 files that are created? Is "outdata1.csv" empty? 

Hope that helps!

Cheers,
matt.

~~~~~~~~~~~
Matthew Blackwell
Institute for Quantitative Social Science
Department of Government
Harvard University
url: http://www.mattblackwell.org

On Thursday, March 29, 2012 at 2:35 PM, Ringelhan, Stefanie wrote:

Hi,

 

I’m rather new to Amelia and have some troubles with it. Thus I would be very happy, if you could give me a hint. Here are my problems I’m dealing with:

I have two warning messages coming up when I try to run the imputation command on Amelia (using R) and I don’t know how to handle them:

 

·         “You have a small number of observations, relative to the number, of variables in the imputation model.  Consider removing some variables, or reducing the order of time polynomials to reduce the number of parameters.”

è  What would be a good relation of observations to variables? Right now I have 292 observations and 121 variables in the data file. And I do not want to delete more variables, as they are pretty much the fundamental variables I want to do calculations with. 

·         “The resulting variance matrix was not invertible.  Please check your data for highly collinear variables.”
è  Is Amelia not running properly in the case of collinear variables? Do I have to delete highly correlating variables or is there a way to work around? There are some variables that are very likely to correlate with each other, however I do not want to delete them from the sample as they will be important in further analysis.
 
Furthermore Amelia imputated 5 times, however data was only printed out in the last one of 5 excel sheets (using the command: write.amelia(obj=a.out, file.stem = "outdata", format = "csv")).
 
Do you have any idea why there is only data written to the last excel sheet? Or is this already the merged data sheet from all 5 imputations?
 
Cheers,
Stefanie

 

 

Follow us on Facebook: www.facebook.com/TUM.LSO
__________________________________

Beschreibung: https://xmail.mwn.de/owa/14.1.287.0/themes/resources/clear1x1.gifDipl.-Psych. Stefanie Ringelhan

Lehrstuhl für Strategie und Organisation

Prof. Dr. Isabell M. Welpe

Fakultät für Wirtschaftswissenschaften
Technische Universität München
Leopoldstrasse 139
D-80804 München (Germany)

Tel.:      +49-(0)89-289-24824
Email:   s.ringelhan@tum.de

 

http://www.strategie.wi.tum.de

 

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