|
Activity Number:
|
405
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #301942 |
|
Title:
|
Restricted Heteroscedastic IV Estimators: A Simulation with Small Samples
|
|
Author(s):
|
Luis Frank*+
|
|
Companies:
|
University of Buenos Aires
|
|
Address:
|
Av. San Martin 4453, Buenos Aires, C1417DSE, Argentina
|
|
Keywords:
|
instrumental variables ; IV estimator ; restricted models ; linear models ; regression analysis ; small samples
|
|
Abstract:
|
Given y = Xß+e subject to Rß = r, the paper proposes a restricted least squares instrumental variables (IV) solution (biv*) and simulates it with small samples using as instruments a monotone transformation of the columns of X. The transformation sets to zero log transformed elements x(ij) when x(ij)tends to +0 (situation which arise frequently under a Cobb-Douglas demand function specification) or to their rank otherwise. Preliminary results show that the proposed estimator performs fairly well (in terms of unbiasedness and efficiency) in very small samples but becomes increasingly biased as the sample size or the percentage of zero-elements grow. The simulation is performed by a "feasible" version of the restricted estimator which solves some numerical drawbacks analogous to those in GLS estimation. The biv* estimator is useful to handle other types of not "well behaved" data.
|
