Error Vector Calibration Function

I have a series of error vectors determined by comparing observed values to known reference values. I would like to fit a calibration function to these vectors so I can remove systematic bias from future observations. What's the best way to do this?

Thanks!