I run the Allan variance analyzation on icm4x6xx accelerometer.The z axis white noise and bias instability is 0.0147124 and 0.00143468 respectly,that's 10x higher than y axis and x axis. And the shape of z axis Allan variance is weird(pleace check the attachment). Is there any way to correct this noisy measurement? Or some error modeling about this issue so that I can complementary reduce this error by fusing other sensor.