Duh! Ok, maybe some didn't know this. Anyway, I've done some sensitivity analysis on my model by changing one variable at a time by 10%, and looking at the resulting percentage change in EPS.
First, I simply adjusted the variable by 10% and froze all the other variables, noted the EPS change, and moved on to adjust the next variable (resetting the previous one to how it was originally). Here's a couple of charts for that (click to enlarge):