Question: Do a Monte Carlo simulation. Take the model Y Xe with E[Xe] 0 where the parameter of interest is exp().
Do a Monte Carlo simulation. Take the model Y Æ ®Å X¯Åe with E[Xe] Æ 0 where the parameter of interest is µ Æ exp(¯). Your data generating process (DGP) for the simulation is: X isU[0,1], e » N(0,1) is independent of X, and n Æ 50. Set ® Æ 0 and ¯ Æ 1. Generate B Æ 1000 independent samples with ®. On each, estimate the regression by least squares, calculate the covariance matrix using a standard (heteroskedasticity-robust) formula, and similarly estimate µ and its standard error. For each replication, store b¯, bµ, T¯ Æ
¡ b¯¡¯
¢
/s
¡ b¯
¢
, and Tµ Æ
¡bµ¡µ
¢
/s
¡bµ
¢
.
(a) Does the value of ® matter? Explain why the described statistics are invariant to ® and thus setting
® Æ 0 is irrelevant.
(b) From the 1000 replications estimate E
£ b¯
¤
and E
£bµ
¤
. Discuss if you see evidence if either estimator is biased or unbiased.
(c) From the 1000 replications estimate P
£
T¯ È 1.645
¤
and P[Tµ È 1.645]. What does asymptotic theory predict these probabilities should be in large samples? What do your simulation results indicate?
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
