1. If Y_t = c_o + c₁ X_t + c₂Z_t + e_t , t = 1, 2, . . . , T

X_t and zero mean e_t are stochastically independent, and e_t = e_t-1 + u_t

where 0, u_t is mean zero i.i.d., then GLS can be performed using

a. generalizing the covariance matrix of u_t and then applying OLS.

b. repeated OLS

c. estimating r and transforming Y_t , X_t , Z_t using this estimates

d. estimating u_t and transforming Y_t , X_t , Z_t using this estimates

2.

If Y_t = c_o + c₁ X_t + c₂Z_t + e_t , t = 1, 2, . . . , T

X_t and zero mean e_t are stochastically independent, and e_t = e_t-1 + u_t .

To test Ho: = 0 in above, Durbin-Watson d-statistic gives 2.65, and at 5% significance level, T = 90, k = 3, how do you conclude?

a. Accept Ho

b. Reject Ho , accept negative autocorrelation

c. Reject Ho , accept positive autocorrelation.

d. Inconclusive on Ho

3.

Suppose Y_t = c_o + c₁ X_t + c₂Z_t + e_t , t = 1, 2, . . . , T and e_t satisfies the classical conditions. However, in a regression, Z_t was omitted. If Z_t = r Z_t-1 + u_t

where r 0, and u_t is i.i.d., the D-W d-statistic in the above is likely to be

a. different from 2

b. close to 2

c. close to 0

d. cannot be computed.