In the ADIS16203 datasheet is a figure explaining tilt plane error (reproduced below). In the datasheet it says that 10 degrees of tilt plane error would produce +/-1 degree of inclination error. What if my tilt plane error is greater than 10 degrees? Is there a way to calibrate it out? If it can't be calibrated out, would the next best solution would be to use a 3-axis accelerometer and do the calculations externally? Do you have a document describing how to do these caluclations?

The culprit is trigonometry ;-)

The gravity vector is alway pointing towards the center of the Earth. If you look at any give axis, the reflection of the gravity vector on that axis is 1g*SIN(theta) where theta is the angle formed between the accelerometer axis and the surface of the Earth. So when the accelerometer axis is parallel to the surface of the Earth the acceleration it "sees" is:

1g*SIN(0°) = 0

When you move the accelerometer such that it forms a 1° angle between it and the surface of the Earth is:

1g*SIN(1°) = 17.5mg

So near parallel to Earth you have a change of acceleration of about 17.5mg/° of tilt.

However as you approach close to 90° (parallel to the gravity vector) there is very little change. A 1° change at say 80° results in:

1g*SIN(81°) - 1g*SIN(80°) = 987.7mg - 984.8mg = 2.9mg/° of tilt.

Using a 3 axis accelerometer you can always choose to use an axis that is closer to parallel to the Earth so as to keep the sensitivity to inclination change high in all positions. Application note AN-1057 explains that.