Hi tr1. You are right about some of what you say. The degree of confidence that can be placed in a trend depends on the variability within the data (variance) and the sample size. The bigger the sample size the smaller the correlation (r) needed for it to become statistically significant. Statistical significance is measured in units such as .10. .05, .01 etc signifying that (for those examples) there is a 1 in 10, 1 in 20 or 1 in 100 chance of the result having occurred by chance. With a sample size of 10, the r of .566 would be significant at .10 in a 2-tailed test, and at .05 in a 1-tailed test. A 2-tailed test would be appropriate here, as there are no theoretical reasons to expect the result.
So to sum up, there is a small but real possibility this result is meaningless. Using daily data would have been better, as it would give more observations, but I don't have that data.
And you are dead right about correlation not equalling causation.
