The IIS3DHHC is a triaxial accelerometer that outputs 16-bit data, 3-axis, ±2.5 g acceleration data.
So (5g)/(2^16 bits) = 7.63*10^-5 g's/LSB.
arctan(7.63*10^-5) is essentially the same thing (small angle approximation in radians)
and that converts out to 0.00437 degrees -- which is about 0.2622 arc-minutes -- which is about 15.74 arc-seconds.
So the LSB represents about 16 arc-seconds of resolution -- is that correct?
The other way I can talk myself into a solution is to eliminate 1 bit for the sign -- leaving 15 bits for a full scale value. As an inclinometer, I'd never expect the value to exceed 1g -- so now we're down to 14 bits corresponding to 1g.
At the same time, I see that there is a 2%FS non-linearity -- thats 10-11 bits of data. Where is the deviation the greatest, and how does that affect the usefulness of this device as an inclinometer? By my count -- I'm down to 3-4 bits of reliability.