Significance of correlation coefficients
Posted: July 28th, 2009, 4:03 pm
Let's assume I have two time series, with given number of degrees of freedom, of independent normal random variables.Is it possible to generate is a 3-D graph showing:x = sample correlation, r, between -1 and 1y = actual correlation, p, between -1 and 1z= probability of sample correlation being x given actual correlation of yFor example if I wanted to know: is the probability that (i) the sample correlation is 0.2 given the actual correlation is 0 greater than (ii) the probability that the sample correlation is 0.9 given the actual correlation is 0.7? then I could read off the graph (0.2,0) and (0.9,0.7) etcWould love to hear from any stats experts out there.Cheers,Jon