What do we use to make a prediction when there is not enough evidence to conclude that there is a significant linear relationship? Do we use the mean of the y values or the regression equation?