Inference in Linear Models with Nonstochastic Biased Factors | ||||
The Egyptian Statistical Journal | ||||
Article 4, Volume 40, Issue 2, December 1996, Page 172-181 PDF (6.96 MB) | ||||
Document Type: Original Article | ||||
DOI: 10.21608/esju.1996.314788 | ||||
![]() | ||||
Author | ||||
Abdul-Mordy H. Azzam | ||||
Faculty of Commerce, Alexandria University, Egypt. | ||||
Abstract | ||||
Obenchain (1977) claimed that ridge techniques with nonstochastic of biased factors don't generally yield "new" normal theory statistical inference than that used in least squares technique, and that the t and F statistics are identical under both techniques. Theorems (1)-(3), in this paper, prove that this is true when using the unbasid ordinary least squares estimator S2 of σ2 Moreover, a counter example is introduced to show that the normal theory doesn't apply when using the ridge regression estimator Sr2 of σ2 instead of using the least squares estimator S2. | ||||
Keywords | ||||
Ordinary Least Squares (OLS); Sum of Squares of Errors under Least Squares (SSEols); Sum of Squares of Errors under Ridge (SSEr); Best Linear Unbiased Estimator (BLUE); Uniformly Minimum Variance Unbiased Estimator (UMVUE); Mean Square Error (MSE); Canonical Parametrization | ||||
Statistics Article View: 83 PDF Download: 9 |
||||