Suppose we compute the regression through the origin of y on x. (That is, we use OLS to estimate y = β1x + u.) Assume that, in our sample,
statistics
Description
1. (6 points) Suppose we compute the regression through the origin of y on x. (That is, we use OLS to estimate y = β1x + u.) Assume that, in our sample,
(Note: By the Cauchy-Schwartz inequality, we know the expression on the left-hand side can’t
exceed unity). Assume also that x = 0.
Show R2 ≡ SSE/SST will exceed unity if the absolute value of the sample mean of y exceeds
the sample standard deviation of y.
(Hints: use SSE =
Pn
i=1(ybi − yb)
2
), and the fact that the OLS estimate of β1 will be βˆ
P
1 =
n
i=1 P xiyi
n
i=1 x
2
i
2. (5 points) Suppose that assumptions MLR.1-MLR.3 hold but we replace MLR.4 with MLR.4’:
E(u|x1...xk) = 5. Show that the OLS estimators of the slope coefficients are unbiase
3. (6 points) Suppose Y ∼ t(m). Use the Law of Large Numbers and the Continuous Mapping
Theorem to show that Y converges in distribution to a N(0, 1) random variable as m → ∞.
(Hint: Start with the representation of a t and a chi-square random variable from the Statistics
Review
