On Fri, 18 Apr 2008, Sven Schreiber wrote:
Ok Jack, thanks for your effort, I appreciate it.
Turns out that the original formula partially uses the wrong
series (the last two), and that during the renaming stage in
order to reduce the line length that error was implicitly
corrected without anyone noticing until now.
So it's not a gretl bug at all!
Sorry for the noise, we really should have cross-checked with a
different software first. But hey, we're all only using gretl
;-)
Thanks, Sven. I had just come to that conclusion myself: somebody
had entered the level rather than the log-difference into the
formula for two of variables, which of course gave you a spurious
upward trend.
BTW, it looks as if someone has done some rather laborious
re-basing of the price series by hand. Here's the preferred way:
open d_ppi100rpiv.gdt
loop foreach i Fruits___melons..Other_miscellan
sprintf sername "%.12sreb", "$i"
series @sername = $i/$i[2000:01]
endloop
Allin.