relativeVariance property
Noise variance of this reading relative to the model's measurement
variance, so the effective noise is
relativeVariance * model.measurementVariance.
Use 0.5 for a reading you trust twice as much (an average of two
weighings, say), 4.0 for one you trust half as much, and a large value
to set aside a reading you believe is a mistake. Must be finite and not
negative; zero asks for the reading to be matched exactly.
Implementation
final double relativeVariance;