Make sure you have a vertex at each point you want to gather data, and one somewhere on the base.
Add a 'workflow sensitive' sensor to each of these vertexes.
Run your study, create an acceleration plot, then probe the plot using the sensors.
In your example the input is 1g at all frequencies, so no extra math is needed to graph the output to input ratio.
If your input is not a constant 1g then I believe you'd have to save the data out and do the division externally (probably in excel)
Ah alright, I tried "acceleration" sensors instead of "workflow sensitive", this does the trick now.