On Fri, May 26, 2006 11:53, Sven Schreiber wrote:
Dear list(eners),
I'm not sure (yet), but there might be a bug in the histogram
calculation and / or display in gretl 1.5.1. (This is on a German
windows xp.)
If you have the time, look at variable "Dax" (students' forecast of a
German share index) in the attached datafile: the supposed histogram
looks more like a mixture of a real histogram (displaying empirical
density for some chosen bins) and a bar graph (displaying relative
frequency of values).
Or is my observation due to some inadvertent user setting? Also, it
seems to me that when I used that data last semester (with gretl 1.4 or
1.5.0), this didn't happen (although I don't remember exactly and could
be wrong).
Thanks for clarification,
Sven
I'm the one blame for this. Actually, in 1.5.1 there is some code that I wrote
to handle discrete variables. The idea was (and still is) to revamp the "freq"
command and implement the "xtab" command for contingency tables for discrete
variables. This was not advertised at the time of release because it wasn't
really ready (and still isn't) for public consumption, although it should work
from the command prompt.
Unfortunately, Allin and I realised AFTER release that the changes I had made
had some nasty side effects, one of which is what you're seeing. Basically,
"Dax" is treated as discrete rather than continuous. This is fixed now in CVS.
Actually, from the next version, you'll have the option of setting variables
as discrete and steer the way things are done accordingly.
Sorry for that.
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona