|
Conference on Quantitative
Social Science Research Using R
(click on the title to download the abstract)
|
|
H. D. Vinod
|
|
Superior Estimation and Inference Avoiding Heteroscedasticity and Flawed Pivots:
R-example of Inflation Unemployment Tradeoff
|
We use a new solution to heteroscedastic regression problem while avoiding so-called incidental parameters (inconsistency) problem by
using recently discovered maps from time domain to numerical values domain and back. This involves a parsimonious fit for sorted
logs of squared fitted residuals. Dufour (1997) showed that inference based on Fisher's pivot (dividing by standard errors) can be
fundamentally flawed for deep parameters of genuine interest to policy makers. Hence, we use Godambe's (1985) pivot, which is always a sum of T
items and asympotically subject to the central limit thory. We provide R functions to implement the ideas using the Phillips curve trade
off between inflation and unemployment for illustration.
|
|
|
|
