This is for IE7 to hold div open
Why does the test of a correlation (Ho: rho = 0) have N-2 degrees of freedom instead of N-1 degrees of freedom? There's only one correlation being estimated instead of two, so why are two degrees of freedom used?
Remember that estimating the correlation coefficient is a special case of using the simple linear regression model. This regression model takes the form:
y = a + bx + e
where a (the intercept), and b (the slope) are the two parameters in the model to be estimated. Since two values are being estimated, two degrees of freedom are lost.
If you have further questions, send E-mail to stats@ssc.utexas.edu.