1

"Backward variable selection"is a technique for choosing the "best"set of X variables for a regression model by:

	A.	regressing Y on *all* X variables and then removing X variables one at a time that don't contribute to the model by some threshold amount.
	B.	maximizes R2
	C.	regressing Y on no X variables and then adding one X variable to the model at a time until no more X variables improve the model at some threshold level
	D.	sequentially adds and removes X variables in search of the best set of X variables

2

If an X variable is added to a regression modelequation and that X variable does not improve the model at all,the F statistic will:

	A.	increase because more "noise"is added to ESS/df
	B.	decrease because MSS/dfdecreases while ESS/df increases
	C.	stay the same because MSS/dfand ESS/dfchange at proportional rates
	D.	decrease because MSS/dfdecreases while ESS/dfstays the same

3

"No multicollinearity" is an OLS assumption that concerns:

	A.	sphericity
	B.	the partial correlationsof each X variable with the Y variable
	C.	the absence of high correlations among the X variables
	D.	the distribution of error terms