A dependant variable is rgressed on K independant variables, using n sets of sample observations. We denot SSE as the error sum of squares and R^2 as the coeffcient of determination for thisestimated regression. We want to test the null hypothesis that K1 of these indpeendant variables, taken togehter, do not linearly affect the dependant varibale, given thaat the other (K - K1) independant variables are also being used. Suppose that the regression is re-estimated with the K1 independant variables of interest excluded. Let SSE* denote the error of sum of squares and R*^2, the coeffcient of determination for this regression. Show that the statistice for testing our null hypothesis can be expressed as follows:
(SSE* - SSE) / K1 R^2 - R*^2 n - k - 1
---------------------- = --------------- x -----------
SSE / (n - k - 1) 1 - R^2 K1