Accepted:
20 Dec 2003
| |
/vol10/hermos01/ | © Copyright 2002 | |||
| Volume 10 | Received: Accepted: |
16 Jun 2002 20 Dec 2003 |
|||
|
On multicollinearity and artificial neural networks
Carpio, K.J.E. & Hermosilla, A.Y. |
|
| Abstract | |
| One of the many problems encountered in coming up with a multiple linear regression model is the presence of severe multicollinearity in the data set. In this paper, the focus is on the mathematics of multicollinearity -- what it is, what it does to the model, how it can be detected and combated. Aside from the classical methods, artificial neural networks are also employed as an alternative to combat multicollinearity. Softwares such as Statistical Package for the Social Science (SPPS) Release 7.0 and 10.0 for Windows, MATLAB version 5.3 and Stuttgart Neural Network Simulator (SNNS) version 4.1 are used to carry out the massive computations in analyzing the data of the mathematics grades of the BS Mathematics graduates of the University of the Philippines. | |
