Russell
Posts: 1082 Joined: April 2005
|
The hypothetical made-up data thing is priceless.
Let me make sure I've got this right though; correct me if I'm wrong:
Given a plot of 38, 380, or 38000 (X,Y) data points, even if they all fall on a line, no matter what the correlation coefficient is, you can never have any confidence about the relationship between X and Y, because it's possible those 38, 380, or 38000 were complete flukes, and the next 380000 points will probably, might, for all we know, could conceivably could, with a probability somewhere around 10^-47(*), demonstrate no significant X-Y correlation.
* (A number I pulled out of thin air. Surely it's as valid as any other!
-------------- Must... not... scratch... mosquito bite.
|