When the best and worst lines have different y-intercepts, the uncertainty in the y-intercept is calculated as |best y-intercept − worst y-intercept|.