1

When Weighted Least Squares (WLS) is utilized to correct for heteroskedasticity, the best choice of weight variable is:

	A.	one that is proportional to the (non-constant) variance of the errors
	B.	any X variable that is highly correlated with other X variables
	C.	the inverse of the square of the smallest variable
	D.	a constant vector (with only one value)

2

The dummy variable coding method that uses only 1 and 0 to represent values is

	A.	effect coding
	B.	interaction coding
	C.	reference coding
	D.	forward selection

3

The theory that we can test certain constraints on a regression model by further decomposing the Error Sum of Squares is the:

	A.	general linear hypothesis
	B.	theory of least squares
	C.	general linear model
	D.	correct model specification

4

The dummy variable coding method that uses 1, 0, and -1 to code the distinct values of a non-numeric variable is called:

	A.	effect coding
	B.	reference coding
	C.	SAS coding
	D.	inverse coding