a very small p-value of a model coefficient does not indicate that its associated x-variable contributes a large amount to explaining or predicting a dependent (i.e. y) variable.