The Problem of Inconclusive Region in Autocorrelation Tests in Least Square Regression | ||||
| The Egyptian Statistical Journal | ||||
| Article 1, Volume 57, Issue 1, June 2013, Page 1-17 PDF (19.1 MB) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/esju.2013.314342 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Naglaa A. Morad 1; Ameera Najem Obaid2
| ||||
| 1Department of applied statistics and econometrics, Institute of Statistical Studies and Research, Egypt | ||||
| 2M.Sc Student | ||||
| Abstract | ||||
| The Durbin-Watson (DW) test is one of the most widely used tests for autocorrelation in regression models. The DW test has, however, an important limitation: the test is indeterminate when the test statistic falls into the so-called Inconclusive (Ignorance) Region. The Inconclusive Region exists because the true distribution of the DW statistic is not tractable. Though there have been suggested a number of approximation methods to establish more accurate critical values for the DW test. This paper compares between the three tests of Inconclusive Region of DW test using a Monte Carlo study, these tests are Theil-Nagar (1961) (TN), Henshaw (1966) (H) and Durbin-Watson (1971). This paper concludes that H considerably outperforms the original DW test and identical to Durbin-Watson (1971) approximation. | ||||
| Keywords | ||||
| Autocorrelation; Durbin-Watson test; Monte Carlo simulation | ||||
|
Statistics Article View: 148 PDF Download: 24 |
||||
