Hi all,
I am working on a Mac and hence can't use Amelia View.
For configuration ease, I set up the configuration for an imputation
routine with Amelia View on an old Windows laptop which is now
powerful enough for the calculation. Hence, I need to adapt the R
archive file to run it on my Mac.
When I look into the R archive file created by Amlia View, I wonder
how the case priors (the similarity matrix between cases) are coded
for to be understood by Amelia.
R lets you transform a vector into a matrix by indicating the vector,
number of lines and columns. With my 64 cases (cross-sectional
units), this should produce a vector with the length of 64*64=4096.
Amelia View displays only half of the matrix of n*(n-1)/2 (if I am
not mistaken) to avoid redundancy. How are the redundant fields coded
(do they receive a 0 and Amelia knows that they are redundant values?)?
What made me struggle was that the vector created by Amelia View is
too short to represent the n^2 matrix (it was around 4014 instead of
the 4096 in length and had a few bizarre non integers at the end). So
the question, how does one code the case priors matrix in a R script?
For the priors there is no ambiguity since they consists of a n*n and
not of a n*(n-1)/2 matrix.
I have on additional question: how does one indicate in the R script
that the output is to be written as a .dta ?
Thanks for your help,
Raffael
__________________________________
Raffael Himmelsbach
Institut d'Etudes Politiques et Internationales
Batiment Anthropôle, Bureau 5128
Université de Lausanne
CH-1015 Lausanne
Phone: +41 21 692 31 76
Cell: +41 79 756 85 10
Show replies by date