The Exact Distribution of a Least Squares Regression Coefficient Estimator After a Preliminary t-Test

The specification of a regression model is often determined in part by testing the significance of one or more potential regressors, and then including them in the model only if they are apparently significant. This paper considers such a preliminary-test strategy in a two-regressor model when a single t-test is conducted. The exact sampling distribution of the preliminary test estimator of a regression coefficient is derived and illustrated, and some implications for confidence levels are explored.


Issue Date:
Sep 09 1990
Publication Type:
Working or Discussion Paper
Language:
English
Total Pages:
23




 Record created 2017-09-11, last modified 2017-09-11

Fulltext:
Download fulltext
PDF

Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)