On Mon, 10 Sep 2018, Allin Cottrell wrote:
I agree it would be nice from the user's PoV if "join"
could handle this
case. However, my first reaction is that this sort of transposition and
reshuffling is not in join's present repertoire and would induce a fair
amount more complexity, plus it means adding at least three more options to a
command that already has more than any other.
I agree on this.
Just thinking out loud, but maybe we could have a pstack() function
(a proper
function) that does the business of transposing and reshaping the data, with
"skip_missing" temporarily turned off. It could take two arguments: the
integer number of individuals and an optional list argument (defaulting to
the entire dataset), and would return a matrix with panelized variables in
the columns.
In itself that would leave the user with the task of extracting the series
from the matrix. But maybe this pstack, or whatever we call it, could write
out the transformed dataset as CSV (or gdt) so it could then be loaded easily
-- either in addition to, or instead of, returning a matrix.
Thi doesn't have to be in libgretl. In fact, it could be part of the
"extra" package, unless I'm missing something.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------