Michael Daly wrote:
On 16-Jun-2005, Peter wrote:


boats that are 18' long overall will almost
always have waterline lengths greater than boats that are 14'


Fine, but we were comparing kayaks that were only a foot and a half or
so different in length.

Your previous statement: "there is no correlation between overall length
and waterline length in kayaks" made no such distinction that it only
applied to some set of kayaks that all had about the same length, nor
was it limited to sea kayaks.

Of the 105 kayaks on the web page of Sea
Kayaker data, the average length is 5.2m (17 ft) with a standard deviation
of 41cm (16 in). 78% of the kayaks fall within one standard deviation of
the mean length. We're not talking about huge differences in length
typically, especially since the standard deviation is comparable to the
differences in LOA and LWL.


but it is very high (correlation coefficient is probably around 0.95).


Instead of pulling these numbers out of your ass, how about some facts?

Based on the data I posted on 18 kayaks (showing percent differences
in LWL and LOA), the actual correlation coefficient is 0.79.

Naturally the correlation coefficient will be less if you restrict the
kayaks under consideration to ones with fairly similar lengths (all but
one in the range from 16' to 19'). In a more complete list with play
boats, WW boats, surfskis, etc. also included the coefficient would be
much higher. Since your original statement just referred to the general
category "kayaks" my estimate was based on this broader selection.

However, a correlation coefficient of 0.79 is a far cry from your
original claim that there is "no correlation" which would imply a
correlation coefficient of 0. The numbers in this case are much closer
to perfect correlation than they are to no correlation.

In the reference to statistical terms I cited earlier, any correlation
coefficient of 0.5 or higher is regarded as "high" (0.1 - 0.3 is small,
0.3 - 0.5 is moderate) and greater than 0.7 is "very high."

Not exactly
tight.

Even taking your specified subset of kayaks, the correlation is "very
high" rather than your original statement that it is nonexistent.