On Fri, 16 Dec 2005, Sven Schreiber wrote:
I have just downloaded the new version, and I'm impressed.
Thanks!
First of all, it seems that the German interface is only partially
complete, being a mish-mash of English and German. So I'm
volunteering to complete the translation (or maybe have it
completed...), if somebody tells me how to do that.
The previous translator, Markus Hahn, did not quite finish and has
not updated the translation lately, so there is work to be done on
that. The "official" reference for translation of gretl is
http://www.iro.umontreal.ca/translation/registry.cgi?domain=gretl
I'm not quite sure what the procedure is for switching from one
translator to another. Another possibility is this: join up as a
member of
sourceforge.net, and I can then give you translator access
to gretl CVS. Then you can update gretl's "de.po" file directly.
From my point of view this makes things easier than working via IRO.
(Although it's good to keep IRP up to date too.) Let me know what
you decide.
Next, over here we teach the cumulative distribution function
(cdf), the Lorenz curve, and the Gini coefficient as part of
introductory statistics. No doubt others do as well, yet afaik
they're not implemented in gretl. So I've written a little script
to calculate the needed bits (but only to a first approximation:
for example the data are not collected into bins as the
corresponding histogram in gretl does); maybe the functionality
could be added to a future release?
Thanks for the suggestion. Yes, I think this should be quite easy
to add as a built-in gretl command (if you have suggestions
regarding the user interface for that, please say). In the
meantime, here's a reworking of your Gini script as a gretl
"user-defined function", which makes it a little more flexible.
<script>
function gini_func clear
function gini_func (series x)
series sx = sort(x)
scalar n = nobs(sx)
scalar sumx = sum(sx)
genr index
scalar gini = 2 * sum(index*sx) / (n*sumx) - (n+1)/n
series cdf = index / n
series lorenz = cum(sx) / sumx
printf "gini = %g\n", gini
printf "gininorm = %g\n", gini * n /(n-1)
return scalar gini, series cdf, series lorenz
end function
# Ramanathan academic salary data
open data8-1
(gini, cdf, lorenz) = gini_func SALARY
print -o cdf lorenz
lgrf <- gnuplot cdf lorenz cdf --with-lines
lgrf.show
</script>
By the way, in the course of writing this I noticed a bug in the
assignment of multiple return values from user-defined functions.
This is now fixed in CVS, but the workaround is to ensure (as in the
example above) that the order in which variables are declared within
the function is the same as the order in which they appear in the
"returns" listing.
Finally, and totally unrelated to the previous points, it would be
nice if seasonal dummies could be added to ADF tests and VAR/VECM
systems automatically, i.e., by just clicking a check-box.
Otherwise, for unadjusted data one always has to do the tests "by
hand"
Seasonal dummies can be added via check-box in the case of
VARs/VECMs in gretl 1.5.0 (provided the data are recognized as
seasonal). As for the ADF test, that's a possibility. But I wonder
what that would do to the distribution of the test statistic?
Allin Cottrell