16 Mar
2004
16 Mar
'04
10:45 a.m.
Hi,
I got this email list address from Gary King and hoping that anyone in the
group might be able to help with Amelia.
I have installed the program, input the file, defined the output file,
defined the AMfully, AMidvar, AMords and tried to run the multiple
imputations. But, I get the following error message: "Insufficient
observations for estimation, use prior or increase prior already
implemented."
I have set AMempri from -2 to +2 - but still get an error message, then I
have set it to +3, but go the message: "If _AMempri>0, user should not set
any hyperparameters of the prior."
This is where I had to stop my efforts, because I don't know what to do
next. I tried to find hints in the Amelia handbook, but wasn't lucky...
Any advice is more than welcome!
Thanks a lot!
Ines
----- Forwarded by Ines Mergel/FS/KSG on 03/16/2004 10:33 AM -----
king(a)harvard.edu
To: ines_mergel(a)harvard.edu
03/13/2004 11:16 cc:
AM Subject: Re: Amelia question
Please respond
to king
I'd install Amelia in a directory with no spaces. so c:\amelia or
c:\programs\amelia should work. if it still bombs, then you'll need to
adjust the memory reserved in the configuration file.
i suggest you get that working and try your problem with a small number of
variables. once its working, increase the number of variables and see how
it goes. Amelia should work for you unless you have a lot of variables.
The number of parameters in the problem increases fast in the number of
variables.
If you'd still like to find someone to help, send email to Gov 2001
<gov2001-l(a)fas.harvard.edu>du>, which is a class I taught last year. Anyone
in that class would do a great job.
Best of luck with your research,
Gary
: Gary King, King(a)Harvard.Edu http://GKing.Harvard.Edu :
: Center for Basic Research Direct (617) 495-2027 :
: in the Social Sciences Assistant (617) 495-9271 :
: 34 Kirkland Street, Rm. 2 HU-MIT DC (617) 495-4734 :
: Harvard U, Cambridge, MA 02138 eFax (928) 832-7022 :
On Sat, 13 Mar 2004 ines_mergel(a)harvard.edu wrote:
Hi Gary,
I am research fellow at the National Center for Digital Government and
working together with David Lazer who is also supervising my dissertation
project on the diffusion of e-learning practices from a social network
perspective. I have conducted a survey in a Swiss Business School and got
86% response rate, which included also respondents who refused to fill
out
the survey, but will be included in the network
analysis because they are
nominated by other actors in the network.
I noticed that I have a lot of missing data in a very small data set. I
used the impute command in Stata and the program deleted a lot of cases
so
that I endes up with only 47 valid cases (total number
of cases: 98).
Also,
I heard that Stata is not able to do a multiple
imputation but is only
doing a one time regression imputation. I was told that I might be able
to
use Amelia, but I am not sure if my dataset is big
enough for the
program?
Are there any limitation?
I have installed Amelia on my machine in a very short path: C:\Program
Files\Amelia and created a shortcut on the desktop, but when I try to
start
the program it only opens the window very shortly
before it disappears. I
read in your website, that this can happen, do I have to install the
program directly on c:\?
Also, is there maybe someone at CBRSS who would be able to assist me with
the imputation process? I sbscribed to the Amelia list serve, but I am
afraid that my questions might be too basic?
Thanks a lot for your help!
Ines
- - - - - - - - - - - - - - - - - - - - - - -
Ines Mergel
Research Fellow
National Center for Digital Government
John F. Kennedy School of Government
Harvard University
79 J. F. Kennedy St., T363
Cambridge, MA 02138
Office phone: 617-496 6166
Office fax: 617-496 1722
- - - - - - - - - - - - - - - - - - - - - - -