[R] Gini with bias correction

[David Winsemius]

On Oct 26, 2012, at 9:32 AM, Diego Rojas wrote:

> Hey there,
> 
> I was wondering if someone could tell me if there's a package or command
> that allows me to compute a GINI coefficient using a vector of weights.
> Also the coefficient should be  bias corrected.

I do not think you have done your "homework". This is the blindingly obvious search strategy on Rseek:


http://www.rseek.org/?cx=010923144343702598753%3Aboaz1reyxd4&q=gini+coefficient+weighted&cof=FORID%3A11

You could also use the same search strategy with findFn in sos:

> require(sos)
> findFn("weighted gini coefficient")
found 13 matches;  retrieving 1 page

Downloaded 10 links in 7 packages.

After you have gone through that process,  the issues of bias correction may still be open, but you will need to explain what sort of bias correction you envisage. I know the rms/Hmisc package provide validation and optimism bias estimates based on a bootstrapping procedure, and that their author's position has been that using AUC or its equivalent the Gini coefficient as a model selection criterion is statistically less powerful than regression methods.



David Winsemius, MD
Alameda, CA, USA