> "Gartland, Myles" wrote:
>
> I am not sure I understand the concept of a "significant" correlation.
>
> To me a correlation is just a mathematical relationship between two
> variables. What makes a significant correlation?
A correlation can also be between two sets of values derived from a
*sample*.
In such a case, its difference from a specified "null" value, usually
0, is "statistically significant" if very few samples (of that size)
from a population with the null-value correlation would differ as much
(or more) from the null value as the sample does.
This is not a fantastically useful concept because it depends on the
sample size. For small N very few correlations are statistically
significant; for very large N most of them are. As the sample size is
usually unrelated to the population, "statistical significance" is not
an attribute of the population.
-Robert Dawson
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
