Philippine E-Journals

Model Subset Selection Using Variance Inflation Factor (VIF) in wavelet Part l · Multiple Regression

Noel P. Cabigon | Bhavik R. Bakshi | Servillano Sb Olano Jr.

Discipline: Engineering

 

Abstract:

This criterion variance inflation factor (VJF) in wavelet domain is used to identify the model subsets that provide an efficient procedure for discriminating variables. The discrimination method is based on two principles: first, the collinear variables' VIF increases with scale while the independent variables do not show significant changes; and second, the variance of the collinear variables VIF is very large compared to that of the independent variables. The procedure is applied to a model that has twenty explanatory variables but with single response variable. The algorithm can be applied to any family of linear models that form a lattice of multiple linear models.