On Mon, February 26, 2007 04:14, Al Leong wrote:
We are teaching courses on nonparametric and panel
econometrics and wonder how/if the following are
possible in Gretl:
kernel density estimation
you can get this in the GUI client under "Variable/estimated density
plot". Not possible yet with the CLI interface.
3-D kernel density plot (can be manipulated as in Matlab)
Not yet. However, you can use R, which is great at these things. Gretl
can export data to R quite easily. Example:
open mroz.gdt /* supplied among gretl's examples */
Tools-> Start GNU R
in R, do
library(np)
foo <- npcdensbw(formula=HA~WA)
npplot(foo,phi=30,theta=120,view=c("fixed"))
and you should see a nice graph of the bivariate density of husbands' and
wives' ages. Please refer to the documentation of the R package "np" for
more info.
bootstrapping (case-based and residual resampling)
There's a section in the user's guide entitled "Resampling and
bootstrapping". Have a look at that.
nonparametric regression
Again, not yet, but R is your friend.
panel co-integration tests (Kao test and Pedroni test)
This, I believe, can be done in a script: essentially, these tests
involve a possibly large number of OLS regressions (maybe within a
per-unit loop) and little more. Have a look at the manual chapter on
panel data and at the "Panel data specifics" section in the chapter
"Special functions in genr".
Please note that, in the version of gretl that is about to be released
any day now, a few things have changed in the way panel datasets are
handled, so be sure to run an updated version.
Hope this helps.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
FacoltĂ di Economia "G. FuĂ "
Ancona