Do you know how to prove using homology theory that there is no retraction from the n-disk to its boundary? It is the same here:
Suppose such a function f exists. Call g the restriction of f to the northern hemisphere of S^n which we will regard as the n-disk D^n. Call h the restriction of g to E. By hypothese, h is a diffeo and so has degree ±1. The following diagram commute:
D^n<---E
|...|
|g...| h
|...|
E<-------|
that is, g o i = h where i is the inclusion of E in D^n. Passing to the realm of (n-1)-degree homology, the above diagram becomes the following comutative diagram
0<------Z
|...|
|...| ±1
|...|
Z<-------|
which is absurd.