Hey ianchenmu.

If you have a root, then the inverse function would set x = 0 and y = the root. Now if you can find where this happens with the inverse function theorem and do an approximate taylor expansion with a linear component, then you can estimate the changes in the roots.