7 Sep
7 Sep
6:20 p.m.
one way is to transform the data to an unbounded scale. e.g., if all
observed and unobserved values of a variable are positive, then transform
the positive ones by taking the log. then impute with Amelia (on the
unbounded log scale). then transform the resulting imputed and observed
values by taking the exponential.
you can also use the priors in Amelia II to help as well.
Gary
---
Gary King
David Florence Professor of Government,
Director, Institute for Quantitative Social Science
Harvard University, 1737 Cambridge St, Cambridge, MA 02138
http://GKing.Harvard.Edu, King(a)Harvard.Edu
Direct 617-495-2027, Assistant 495-9271, eFax 812-8581
On Fri, 7 Sep 2007, Bobbie Mircheva wrote:
Hi,
I have a big panel dataset on countries in Sun-Saharan African with
information on trade and infrastructure between 1990 and 2005.
Unfortunately some of the infrastructure data is missing and I was using
Amelia to fill in the gaps. However, the program returned a bunch of
negative values. How can I fix this?
Thanks a lot
Bobbie
- Amelia mailing list served by Harvard-MIT Data Center [Un]Subscribe/View
Archive: http://lists.gking.harvard.edu/?info=amelia
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.gking.harvard.edu/?info=amelia
8 Sep
8 Sep
3:39 a.m.
use each data set as if it contained fully observed data, and then take
the 5 estimates and combine them as described in our APSR paper (which you
can find on our web site). Clarify in Stata will do this automatically if
you use one of its models, or Zelig in R will do it for a much larger and
growing range of models (both are on my web site). But its easy to do by
hand too. Gary
On Fri, 7 Sep 2007, Bobbie Mircheva wrote:
Gary,
Thanks a lot for the quick response. I did what you suggested and it worked great. I
do have one more question if you do not mind. Since I know have 5 output, which one
am I supposed to use. Or is there a way to compile them? I am working with STATA and
tries to I tried to use Ken Schieve's program, but somehow could not manage to make
it work. I still need to do more manipulations with the data (take first differences,
etc) before I start running regressions.
Any suggestions would be greatly appreciated.
Thanks again!
Borislava Mircheva
PhD Candidate
Department of Economics
American University
https://eagle1.american.edu/~bm6265a/
-----Gary King <king(a)harvard.edu> wrote: -----
To: Bobbie Mircheva <bmircheva(a)american.edu>
From: Gary King <king(a)harvard.edu>
Date: 09/07/2007 06:20PM
cc: amelia(a)lists.gking.harvard.edu
Subject: Re: [amelia] negative values
one way is to transform the data to an unbounded scale. e.g., if all
observed and unobserved values of a variable are positive, then transform
the positive ones by taking the log. then impute with Amelia (on the
unbounded log scale). then transform the resulting imputed and observed
values by taking the exponential.
you can also use the priors in Amelia II to help as well.
Gary
---
Gary King
David Florence Professor of Government,
Director, Institute for Quantitative Social Science
Harvard University, 1737 Cambridge St, Cambridge, MA 02138
http://GKing.Harvard.Edu, King(a)Harvard.Edu
Direct 617-495-2027, Assistant 495-9271, eFax 812-8581
On Fri, 7 Sep 2007, Bobbie Mircheva wrote:
Hi,
I have a big panel dataset on countries in Sun-Saharan African with
information on trade and infrastructure between 1990 and 2005.
Unfortunately some of the infrastructure data is missing and I was
using
Amelia to fill in the gaps. However, the program
returned a bunch of
negative values. How can I fix this?
Thanks a lot
Bobbie
- Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View
6086
days inactive
6087
days old
amelia@lists.gking.harvard.edu
4 comments
2 participants
participants (2)
-
Bobbie Mircheva -
Gary King