Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: RE: Drawing residuals from a Gumbel distribution

 From "Nick Cox" To Subject st: RE: Drawing residuals from a Gumbel distribution Date Wed, 18 Aug 2010 12:18:43 +0100

```To me residuals are what you calculate given data and a model. Reading just the specific paragraphs from the references you helpfully supplied does not make what you want to do entirely clear, but it sounds more as if you want to simulate using error terms drawn from a Gumbel. You can get what might be called unit Gumbel random deviates by e.g.

gen gumbel = -ln(-ln(runiform()))

and you can scale those and/or shift them relative to zero if you wish. This recipe follows directly from the form of the quantile function, which is just the inverse of the distribution function given explicitly by your second reference.

Some of the programs in -gumbelfit- from SSC and the short paper

SJ-7-2  gr0027  . .  Stata tip 47: Quantile-quantile plots without programming
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N. J. Cox
Q2/07   SJ 7(2):275--279                                 (no commands)
tip on producing various quantile-quantile (Q-Q) plots

touch on related (but not identical) issues.

Nick
n.j.cox@durham.ac.uk

Mirriam Gee

I am working with a Discrete Labour Supply Model (Binomial logit
model). My question is how can I draw a set of residuals from ‘Extreme
value- type 1 (Gumbel) distribution `  as explained in:

http://ged.u-bordeaux4.fr/ceddt114.pdf
Page 7 (last paragraph)