The Standard Error (SE) of Prediction: a Measure of Individual Uncertainties
PRO:
- Is easy to display & to grasp visually
- Supports easy comparisons between predictions
- Allows confidence intervals to be calculated at an arbitrary level of uncertainty
- Everybody knows what it means
CON:
- To be precise, predictive errors must follow an approximately normal distribution with a standard deviation σi = SEi
- To be useful, predictive error should follow an approximately normal distribution with a standard deviation σi ≤ SEi
- It cannot account for non-random error (bias) in predictions
- Nobody knows what it means
22nd European Symposium on Quantitative Structure-Activity Relationships, September 16-20, 2018, Thessaloniki, Greece
By Robert D Clark