Hi Steve,
You should can find my test data from different sheet as shown below. 1600Hz just the Sheet1.
It is a good method to calibrate the sensitivity and offset in +1g and -1g filed. This can help to avoid the cross-axis influence.
Like the offset, the slope (sensitvity) will change a little too.
What's more, even keep the same output data rate, you will find out the offset and sensitivity output will change a little too when you turn on -> turn off -> turn on the part since every part has the repeatability issue.
For the I&I standard part with ceremic package like ADXL203, the change will be much smaller.
For the commercial grade part like ADXL345 with plastic package, it will be a little worse.
For ADXL345, you do not need to change the measurement range no matter what application is, you can set it to full-resolution mode to support +/-16g measurement range, then it can give out the constant scale factor: 3.9mg/LSB no matter you need 2g range, 4g range, 8g range or 16g range.
Please note that scale factor also has the nonlinearity issue, eg: in +/-12g range, the scale factor is 3.9 mg/LSB. When the acceleration reach 15g, near the max measurement range of ADXL345, the sacle factor may change to 3.9*0.95=3.705mg/LSB. This is the nonlinearity influence.
BR
Neil