Multicolinearity Detection by Means of the h-Plot of
the Correlation Matrix Inverse
Guillermo Ramírez, Maura Vasquez, Alberto Camardiel, Betty Pérez & Purificación Galindo
Abstract
Multicollinearity generates imprecision in the estimates
of the coefficients in linear models. We propose to use the h-plot of the
inverse of the correlation matrix to obtain a representation of the linear
relations between the predictor variables. In the resulting plot the variance
inflation factor of each variable and the partial correlation between them area
roughly displayed. In order to illustrate the method it was applied in an
anthropometric study of young Venezuelan swimmers.
Key words: Multicollinearity, h-Plot, Partial correlation,
Variance inflation factor.
PDF (Spanish)