content, the following assumptions of the model were met: lack of
multicollinearity in the predictors; linearity of the relationship between
dependent and independent variables; independence of the errors (lack
of autocorrelation of residuals); homoscedasticity of the errors and normality
of the error distribution.
4.3.1. Lack of multicollinearity in the predictors
Multicollinearity,meaning strong correlation among some independent
variables, leads to problems in understanding the relative importance
of predictors. This may happen because highly-correlated
predictors measure the same concepts.
In both cases (OGP and FOM), the two analyzed independent variables
seem to be slightly correlated: X2 (muscle thickness) goes down
when X1 (back fat thickness) goes up. The Pearson product–moment
correlation factor (r) is rX1,X2=−0.46 (P b 0.001) for the OGP model
and rX1,X2=−0.37 (P b 0.001) for the FOM model, confirming a moderate
negative correlation between X1 and X2. Since this correlation, even
if it exists, is of a medium level, it can be concluded the regression
models do not need any transformation due to multicollinearity.