For the ADIS16003, Analog Devices recomend a single pole filtering. Their signal is output through a 32K resistor, and with a x uF capacitor, you lower the bandwidth of the response.

With or without that capacitor, the inertia signal at every change of tilt (from rest going up, down, or from going up/down to the oposite) there is this "parasitic" signal. I was able to separate the one apearing after moving from rest. For the one in a continuous move with change of direction, I intend to aply a four pole filter. This will introduce a group delay of 35 to 40 msecond (Chebishev, Bessel). Do you have any schematics for an active filter with a lower group delay, like 10-15 msec?