On Wed, 20 Dec 2006, Andreas Karlsson wrote:
I have a few suggestions for improvements of gretl, making it
possible for me to use it in other courses:
1. A possibility to generate random binomial variables, for
example by adding a menu entry Add > Random variable > Binomial
That would be easy enough. Also easy to do manually, for example
* add normal random variable, u
* genr binrand = u > 0
2. For frequency plots of continual data (i.e., histograms), amke it
for the user to
a). Choose bin width
b). Choose number of bins
Not hard at the basic coding level. A bit more time-consuming to
construct a good user interface.
3. For discrete data, make it possible to create pie charts
Pie charts are evil, and are not supported by gnuplot. That one
probably won't happen soon, though obviously it's not hugely
These are the most important improvements. What are the chances of
suggestions being implemented soon?
1 and 2 are candidates for fairly quick attention, though I can't
promise just how soon.
Some other suggestions, not necessary for using in the courses I am thinking
about, but which would be quite useful and nice to have in gretl (in order of
1. Support for Multinomial logistic regression
2. Forward/backward stepwise selection for regressions
3. Direct import of R format, SPSS format and OpenOffice.org
These formats all quite time-consuming propositions. Is there an
open spec for the SPSS format?
4. Quantile regression
5. Ordinal logistic regression
Ordered logit is in current CVS.
The estimators you mention are good candidates for inclusion at
some point, but I don't know how soon -- there are lots of things
to be done and only two of us actually coding on gretl.