[Ml-stat-talks] Fwd: REMINDER:CSML Seminar Series, April 4th at 12:30pm Green Hall, Room 0-S-6

Barbara Engelhardt bee at princeton.edu
Mon Apr 3 11:02:46 EDT 2017

Talk of interest.

---------- Forwarded message ----------

*Michael Stein*
CSML Seminar
Tuesday, April 4th, 2017
Green Hall, Room 0-S-6
**Lunch will be provided**

Title: Some thoughts on the use of extreme value theory for temperature

Most studies of extremes focus on fat-tailed distributions. However,
temperature distributions tend to be thin-tailed. Indeed, when one fits
extreme value distributions to temperature data, one generally obtains
negative estimates of the shape parameter, which implies that temperature
distributions are bounded.  This talk will explore a number of issues that
arise when applying extreme value methods to daily temperature data,
including the use of long climate model runs to estimate properties of
extremes, taking account of seasonality and models for dependence.  I will
also describe a weighted composite likelihood approach to estimating the
upper tail of a distribution that avoids having to pick a sharp cutoff for
which observations are excluded from the estimation.
If you would like to be added to the CSML listserv, please email:
*csml at Princeton.EDU* <csml at Princeton.EDU>

Thank you!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20170403/be76aa5a/attachment.html>

More information about the Ml-stat-talks mailing list