I have been developing an inertial sensor which senses rotation. One performance requirment is for bias stability of 0.03 deg/hr 3-sigma. If the distribution of bias measurements were normal or gausian, then the formula would be simple.
3-sigma = 3 * 1-sigma = 3 * (variance) ^1/2
Unfortunately, the bias measurements look like a sin function with random noise. So, if the distribution is not normal, how do I calculate 3-sigma???