Degrees of freedom

CFA Level II
#1

I seem to be losing a lot of points trying to use the right degrees of freedom… I always screw up between using (n-1) and (n-k-1) df. For example, how do we decide how many DF to use when t-testing an intercept or coefficient? I know RSS=k, SSE=n-k-1, and SST=n-1, but can’t figure out a logical way to connect that to anything else…

#2

You lose one degree of freedom for every statistic you calculate. In a multiple regression, you calculate the intercept and k slopes: that’s k + 1 statistics. Thus, for n observations, you have n – (k + 1) = n – k – 1 degrees of freedom left.

#3

what s2000magician said is correct. If you are using degrees of freedom always use: n - k - 1

#4

what about the correlation test?