>>Eric Bohlman  <[EMAIL PROTECTED]> wrote:

>>>What would be good 
>>>similar examples for distributions that are symmetric but not normal (for 
>>>students who aren't yet sophisticated enough for the Cauchy distribution to 
>>>be a good example)?  

>Radford Neal <[EMAIL PROTECTED]> wrote:

>>This is difficult, since there's usually no reason for a real
>>distribution to by symmetric.  Perhaps errors in measurement of
>>longitude?  

Herman Rubin <[EMAIL PROTECTED]> wrote:

>It was assumed, and as we now know mistakenly, that errors
>of measurement were symmetric. 

This may be true in general, but my choice of longitude rather than
latitude was quite deliberate.  The sign of longitude is an arbitrary
convention.  So why would the errors be asymmetrical one way rather
than the other?  It's possible, of course, if some asymmetrical
measurement method is used, but I expect that there are also methods
that have no reason to err one way rather than another.

   Radford Neal
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to