Yumin,
Aside potential differences in the treatment of 0 income and of ties or
weights (but these issues do not seem to be arising in your problem), the
difference between -rspread- and -inequal- (and -inequal7- or -ineqdeco-)
is a factor (N-1)/N (where N is the sample size):
RSPREADGINI*(N-1)/N = INEQUALGINI
Many alternative formulae exist for the Gini coefficient, and these two
commands use different ones (-rspread- use a covariance-based formula). The
difference is typically negligible when dealing with large sample surveys,
but will show up in small samples.
Hope this helps
Philippe
___________________________________________
Philippe Van Kerm
CEPS/INSTEAD, Differdange, GD Luxembourg
___________________________________________
>>>>>>>>>>>>>>>>
Re: st: Re: how to obtain Gini coefficients - rspread or inequal?
--------------------------------------------------------------------------------
