Re: st: RE: Income distribution

[Austin Nichols <austinnichols@gmail.com>]

Abdel Rahmen El Lahga <rahmen.lahga@gmail.com> :
I would recommend instead avoiding the lognormal distribution, and
stick with Dagum or Singh-Maddala.
To quote from the paper you mentioned:
"The most surprising feature...is the fact that the Stage II
adjustment to the synthetic sample does not always improve the
accuracy of the estimate of  inequality. Indeed, for the Singh-Maddala
and Generalized Beta distributions, the adjustment raises the mean
absolute percentage error...."

On Thu, Mar 4, 2010 at 12:24 PM, Abdel Rahmen El Lahga
<rahmen.lahga@gmail.com> wrote:
> Hmm,
> Unfortunately, it seems that most of the fitted functions, using
> grouped data, are unreliable ,
> i.e., they often produce summary statistics which are very different
> from those obtained from individual micro data.
> I suggest the procedure proposed recently  by Shorrocks and Wan (2008)
> ( Ungrouping Income Distributions: Synthesising Samples for Inequality
> and Poverty Analysis: wider discussion paper no 16 ) which can be
> downloded at
> http://www.wider.unu.edu/publications/working-papers/research-papers/2008/en_GB/rp2008-16/
> roughly  the approach of Shorrocks and Wan fits any chosen
> distribution to grouped data to produce a sample of N observation then
> they propose a second stage of adjustement until that characteristics
> of the generated sample exactly match the reported grouped values. The
> simulated data can then be used to estimate various inequality and
> poverty measure.
> The DASP package of Araar and Duclos contains a ready Stata routile
> that apply such procedure
> DASP can be downloeded at http://dasp.ecn.ulaval.ca/
> HTH
> AbdelRahmen

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/