Hello Pawel,
I just wanted to let you know that this is a bug on our part that I'm
trying to get all patched up today. It only happens when you have one
(nontime) covariate to impute, which doesn't come up super often. I'll
send some mail to the list (and to you) when I get it up and running.
In the meantime, if you have any other variables that could help predict
the missingness you could include that variable in the imputation and it
would also solve this problem. Also, you may want to think about the
imputation model you are using. Multiple imputation helps the most when
the MAR assumption holds and that is when we can predict missingness with
the observed values in the data. If you think that the cross-section and
time can predict missingness really well, then you're fine. However, you
may want to expand your covariates to include good predictors of your
missing variable in order to make the MAR assumption more plausible.
I hope that helps.
cheers,
matt.
On Mon, 23 Apr 2007, Pawel Kranzberg wrote:
Hello,
I have no prior experience with Amelia, and am trying to use it to
impute missing observations in column "ON" in the enclosed "1ON.csv"
file. I use only 3 variables: ON, month2 (as time series index),
source2 (as cross-sectional index). So far I have only managed to get
the following alternative error messages:
Amelia Error Code: 33
The time series and cross sectional variables cannot be transformed.
(when I try to transform "month2" into ordinal and "source2" into
nominal variables)
Error in rowSums(AMr1) : 'x' must be an array of at least two dimensions
(with no variable transformation)
I guess the problem might be caused by the file structure. Could
someone help me out? E.g., by sending me a sample input file on which
Amelia works right?
Many thanks in advance, and best regards,
Pawel
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia