[4eyes] Ethics Data Analysis

[Cha Lee]

I found this a while back. I found it pretty interesting, as the
author does present some good examples. Also it has a pretty good, and
practical explanation on when to exclude/include data points.

In practice, many researchers "massage" their data until it generally
fits their goals. The less (or maybe more) corrupt/base/evil of these
people generally hide their fingers with long winded and logical
sounding explanations of why so and so data should be
excluded/transformed/adjusted. The completely corrupt/naive researcher
will simply omit such explanations and exclude/transform/adjust the
data without informing the reader.  A basic thing to know is when to
do such "massaging." These slides describe when it is appropriate to
exclude data points.

As is with any statistical method, there are debates to the validity
of each technique presented here. Some argue that there is no such
thing as statistical outliers, only errors (resulting from human or
mechanical errors). Some argue that of course there must be a
mathematical method of detecting and removing outliers. I'm firmly on
the fence on this issue, and I tend to sway from one side to the other
depending on what project I'm working on, the deadline, and how
beneficial it is to me (thus is my corruption exposed).

P.S. I am not for or against any of these methods. I just think you
should be aware of these issues if you are in the HCI community and
not already aware.

-- 
Regards,

Cha Lee
PhD Candidate
University of California, Santa Barbara
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ethics Data Analysis.ppt
Type: application/vnd.ms-powerpoint
Size: 194048 bytes
Desc: not available
URL: <https://lists.cs.ucsb.edu/pipermail/ilab-users/attachments/20120712/4b8276bd/attachment-0001.ppt>