Hi,
In insurance world, some variables follow Tweedie distribution. Does any of
the Allstat participants have experience on that? I would like to know how
to make a model with Tweedie error structure. Thanks.
Chenghong
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list
[mailto:[log in to unmask]] On Behalf Of Coleman, Mark
Sent: Thursday, May 18, 2006 3:48 PM
To: [log in to unmask]
Subject: Detecting outliers in non-normal data
Greetings,
I work in an industry -- insurance -- where much of the data does not
follow a Gaussian distribution. For instance, researchers often estimate
models of insurance settlement payments using GLMs under the assumption
that the errors obey a gamma distribution. As a general question, I'm
curious if the familiar methods of identifying outliers, such as the two
standard deviation rule, are applicable to such data. I presume in
general that they are not, but I'm curious to hear any suggestions that
AllStat participants might be willing to share.
Thank you.
Best regards,
Mark
|