A New Approach for Model Selection with Two Qualitative Regressors | ||
| Computational Journal of Mathematical and Statistical Sciences | ||
| Volume 1, Issue 1, November 2022, Pages 63-79 PDF (1.21 M) | ||
| Document Type: Original Article | ||
| DOI: 10.21608/cjmss.2022.274302 | ||
| Authors | ||
| Amr R. Kamel* 1, 2; Abdulaziz A. Alqarni3 | ||
| 1Department of Applied Statistics and Econometrics, Faculty of Graduate Studies for Statistical Research, Cairo University, Giza, Egypt | ||
| 2Department of Basic Sciences, Elgazeera High Institute for Computers and Information Systems, Ministry of Higher Education, Cairo, Egypt | ||
| 3Department of General Studies, Jeddah College of Technology, Technical and Vocational Training Corporation, Jeddah, Saudi Arabia | ||
| Abstract | ||
| In social science and statistical literature, ordinal qualitative dependent variable models have received substantial attention in terms of theory and application. However, linear models with ordinal qualitative regressors have been overlooked. In this paper, we propose an approach to regression model selection with two-qualitative regressors, which will be used to build a technique for selecting qualitative variables based on the mean square of the prediction error (MSEP). Several data sets were simulated using two models, and the technique was found to choose the best model in all cases. The findings show that significant improvements in bias and efficiency may be made when compared to other estimates. To diagnose the type of interaction, certain graphs are provided. The proposed approach achieves better accuracy with reduced MSEP, improved stability, and faster convergence compared to other models. | ||
| Keywords | ||
| Classification Performances; Model Selection; Modeling of Interaction; MSEP; Neighbouring Model; Ordinal Qualitative Variables; Prequential Inference | ||
|
Statistics Article View: 193 PDF Download: 229 |
||
