Comparison of Logistic Regression Model and MARS Using Multicollinearity Data Simulation
| Main Authors: | Wibowo, Ananto, Ridha, M. Rismawan |
|---|---|
| Other Authors: | Badan Pusat Statistik |
| Format: | Article info application/pdf eJournal |
| Bahasa: | eng |
| Terbitan: |
Universitas Muhammadiyah Mataram
, 2020
|
| Subjects: | |
| Online Access: |
http://journal.ummat.ac.id/index.php/jtam/article/view/1801 http://journal.ummat.ac.id/index.php/jtam/article/view/1801/1397 |
Daftar Isi:
- There are several statistical methods used to model the effect of predictor variables on categorical response variables, namely logistic regression and Multivariate Adaptive Regression Splines (MARS). However, neither MARS nor logistic regression allows multicollinearity on any predictor variables. This study applies the use of both methods to the simulation data with principal component analysis as an improvement in multicollinearity to find out which regression has better performance. The result of the analysis shows that MARS is very powerful in modeling research simulation data. Besides, based on the criteria of the number of significant major components, accuracy, sensitivity, and specificity values, MARS has more appropriate performance than logistic regression.