Errors in GEV Analysis of Wind Epoch Maxima From Weibull Parents. (by R. I. Harris, Wind and Structures, Vol. 9, No. 3, p. 179-191, 2006.) Discussion.
Errors in GEV Analysis of Wind Epoch Maxima From Weibull
Parents. (by R. I. Harris, Wind and Structures, Vol. 9,
No. 3, p. 179-191, 2006.) Discussion.
(75 K)
Harris, R. I.; Simiu, E.
Wind and Structures, Vol. 10, No. 4, 399-400, July
2007.
Keywords:
wind velocity; wind effects; hurricanes; weather
effects; statistics; structural engineering
Abstract:
One basic question that needs to be raised in an extreme
wind speed estimation context is whether parent
populations can be clearly identified. To this
discusser's knowledge this is not the case for hurricane
or thunderstorm winds. Whether parent populations can in
general be identified for "straight winds" remains to be
ascertained. The discusser's argument on convergence is
well taken, but it would have been useful for the author
to discuss it in the paper, at least briefly, for the
case of the Type I Extreme Value distribution as well.
Similar convergence issues may arise for this case, and
it is still not clear why convergence concerns
applicable to a Type III distribution estimated by a GEV
analysis would not be equally applicable to a Type I
which, after all, can be regarded as a particular case
of the GEV family of distributions. The author's
arguments are predominantly concerned with extreme value
statistics issues that are
not specific to wind engineering, but rather pertain to
statistical issues of much more general scope.
For this reason, as he has done in the past, the
discusser would suggest that the author submit those
arguments to statistics journals for proper scrutiny by
professional extreme value statisticians, rather
than publishing or offering papers to wind and
structural engineering journals. Judging by the
references listed in the author's paper, so far he
appears not to have done so.
Among the four reasons that, in his opinion, may
motivate the proponents of the reverse Weibull
distribution tail, the author mentions "continual
commercial pressure to reduce wind loads." In fact,
basic extreme load estimates iTom presumed "parents" can
yield unsafe estimates of extremes, as
noted by Isyumov, et at. (2003) in connection with
estimation methods based on up-crossing theory
and still used by some wind tunnel operators. The
discusser believes that the motivation for
adopting extreme wind speed models other than the Type I
distribution stems iTom a genuine desire
on the part of statisticians and wind engineers to
develop models that are appropriate, a desire they
share with the author, who should be congratulated for
his efforts to understand the probabilistic,
statistical, and physical elements of this difficult
issue.
Building and Fire Research Laboratory
National Institute of Standards and Technology
Gaithersburg, MD 20899