Hi all,
I wanted to see the mlogit demo, on OSX 10.4.4, running R 2.2.1,
latest version of Zelig. The following describes the problems I
encountered, and provides solutions for some of them. A solution to
the latest problem would be very helpful.
-----
Task: Install Zelig. (Developer tools are installed, make is
available etc.)
Result: VGAM failed to install. So I downloaded the tarball from the
VGAM pages, and tried to install from the local source bundle. This
failed again with a linking error: "ld: can't locate file for: -
lcc_dynamic".
Notes: Judging from the regular R mailing list it is not unusual to
get fortran linking errors on OSX because of compiler
incompatibility, usually concerning restFP and saveFP symbols. The
solution seems to be to hand "-lcc_dynamic.a" to the linker, which
VGAM duly does. However, the latest version of OSX, can't find it,
though the man page for 'ld' suggests its on one of the paths ld
should check. This manual symbolic linking approach makes it
available, though it's unclear to me how robust a fix it is.
Solution: Trawling the web suggested the following incantation:
ln -s /usr/lib/gcc/powerpc-apple-darwin8/4.0.0/libgcc.a /usr/lib\
/libcc_dynamic.a
Out-of-the-box OSX users will want to prepend this with 'sudo' to run
as root. Credit for this solution (if it really is the solution)
goes to the Octave mailing list members:
http://www.octave.org/
mailing-lists/help-octave/2005/4683
VGAM now loads.
-----
Task: Show the mlogit demo. Call: demo(mlogit)
Result: demo runs the model part successfully and chokes on
visualization - vcd package not available.
Solution: install the vcd package
-----
Task: Make vcd run. Call: library(vcd)
Result: Fails due to absence of colorspace package
Solution: Install colorspace package. vcd now loads.
-----
Task: Show the mlogit demo. Call: demo(logit) again.
Result: demo runs up until the first ternary plot is shown. Then
fails with error message:
ternarypoints(ev.weak, pch = ".", col =
"red")
Error in plot.xy(xy.coords(x, y), type = type, pch = pch, col =
col,
bg = bg, :
plot.new has not been called yet
-----
And that's as far as I've got. Any thoughts on getting further would
be greatly appreciated.
Will
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)latte.harvard.edu
[un]subscribe Options:
http://lists.hmdc.harvard.edu/?info=zelig
Zelig program information:
http://gking.harvard.edu/zelig/