Yes, it should work the same as any other method. Identify some quantity
of interest (a coefficient, predicted vaqlue or something else), and then
run the analysis 5 times and combine the results as per the procedures in
our article.
The problem with high missingness is that your results will be sensitive
to assumptions (variables included, etc) at the imputationa stage. But
this is a feature of your data, for any method that would be applied, not
only thaose in Amelia.
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 Wed, 17 Sep 2003, Thomas Pepinsky wrote:
I'm interested in using quantile regression in a
study of income and wage
effects, and I've used Amelia to simulate 5 datasets (I should note that
I have high-missingness). Does quantile regression meet the assumptions
of Amelia's multiple imputation algorithm?
A second question regards intepretation. Can I combine the results from
my estimates of parameters and variances in the same fashion for estimates
within the quantiles as I can OLS regression and the others?