significant correlation vs. predict better?

["Danni Wang" <liuwang94710@hotmail.com>]

Dear Group,

Hope this question is not far from the field. 
We know a linear predictor could be used to predict
y=Ey+corr(x-Ex)SD(y)/SD(x) from x. When two sequences of data (x_seq,
y_seq) are tested having a significant positive correlation with each
other using non-parametric rank method, is there a way to improve the
prediction of y from x using this information?

Thanks very much,
Danni