On 10/15/2008 12:57 AM, Carlos Rodriguez wrote:
Dear list,
I have been trying to use priors in Amelia with no success. I only
want to set the range as > 0 for the variables that are percentages
because I have noticed in Amelia output datasets that some of the
imputed values are negative. I am not sure how much of a problem the
latter is because, as I understand it, MI is about distributions not
exact values of missing obs. I selected the option "range" in the
priors tab but when I wrote 0 as minimum I got an error message asking
me to include the mean. I only wanted to get positive imputed values
and I have no priors about the mean. Could you please tell me how
this should be done?
if you want to constrain the outputs in this way, you can either round
to [0,1] after imputation or (better) you can transform your variables
going into Amelia as for example ln(p/(1-p)) and then untransform after
imputation. There are some examples of this in our APSR paper.
Second, I have been told that if one plans to run a
(coutnry) fixed
effects model, the country dummies should be included in the dataset
that is loaded in Amelia to perform the multiple imputation. Is this
right or should fixed effects be included in a different way...or not
included at all in the MI but added to each imputed datasets after the
imputation is done?
either way. Amelia will create these dummies for you if you like or you
can enter them yourself if that would be convenient.
Finally, could you please let me know where I can get
some guidance on
the interpretation of Amelia diagnostics. I got all the graphs but I
am not sure what they are telling me or whether overall my MI is ok.
This is a paper that we plan to write but haven't quite gotten to! For
now, the manual is the best source.
Gary
---
http://gking.harvard.edu
Thanks so much in advance for your help!
Carlos
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia