This is a discussion of problems with Joe Mehaffey's
averaging analysis...

First of all there is the graph (which he didn't draw)
which states "very short term averaging shows little
improvement in accuracy due to correlation".  Actually,
the data tells a different story altogether, the real
reason that graph looks like little gain is because
a log has been taken of BOTH the time AND the RMS!

The best way to see whether correlation is creating
"little gain" is to look at the non-log derivative,
something he has failed to provide (largely because
everything on his page was provided by someone else,
and the original graph provider has steadfastly
refused to plot the derivative).

Later on he states "Again, averaging for a period of
a minute or a few minutes gives you very small 
improvements in accuracy", making statements about
"a few minutes" when he only presents John Galvin's
data for ONE minute!  Not only that, but John has
provided the "few minutes" data, but he refuses to
put that on his web page.  Probably because he is
now in the unenvious position of having to claim
that 13-78% (see below) is a "very small improvement 
in accuracy".

reduction caused by 4 minutes averaging (correlation period):
RMS: 16%
Usage Scenario 1 (average error): 13%
Usage Scenario 2 (average search area): 53%
Usage Scenario 3 (maximum error): 30%
Usage Scenario 4 (maximum search area): 78%

The reason behind these actions is because Joe was one
of the camp who were originally claiming that there was
NO improvement from averaging data less than the
correlation period "of 15 minutes".

Furthermore, there's no mention of the different usage
scenarios, except for a brief acknowledgement of the
search area situation, and no mention of the cost
factor, both of which need to be considered for
practical waypoint averaging.