Bus Analytics - Noise

Two P’s of Useful Statistics

●Persistent

●Predictive

Whenever there is human judgment,

there is noise

bias

Difference between the estimator’s expected (average/mean) value and the true value.

noise

Statistical variability of the data in the data set.

Single Measurement Error (SME)

SME = Bias Error + Noise Error

Noise Error is positive when SME is greater than Bias

Noise Error is negative when SME is less than Bias Error

Total Error (MSE)

1.Bias

2.Noise

Such that Mean Square Error (MSE) = Bias² + Noise²

Components of Noise

●Level Noise

●Pattern Noise

●Random (Occasion) Noise

Reducible Uncertainty

What could be known but isn’t at the time of a prediction.

intractable uncertainty

What cannot be known at the time of a prediction

reductible + intractable uncertainty =

perfect prediction

How to Correct Predictions

1.Make your intuitive guess.

1.Find the mean

1.Estimate the diagnostic/predictive value of the information you have

1.Adjust your outside view toward you intuitive guess, to the extent of the predictive value