Inside or Outside the Tested Range: Label Trial-Model Predictions

Calculation is not the same as confirmation

Interpolation concerns predictions within a data domain; extrapolation extends beyond it. A model that fits observed combinations does not automatically remain suitable elsewhere. Predictions in either region require review, with extra caution where the intended condition lies outside the observations.

Primary reference: NIST Engineering Statistics Handbook.

A hypothetical application extension

Suppose an illustrative model was fitted using an approved factor range from 10 to 20. A request at 15 lies inside that range; a request at 30 lies outside it. The software can return both numbers, but the second does not gain evidence merely because the calculation succeeds.

With several factors, checking each separate minimum and maximum may still miss an untested joint combination. Retain the actual observed combinations. The example supplies no HM speed, temperature or pressure range.

Show the prediction domain in the enquiry

Provide factor definitions, units, tested combinations, proposed new condition, model equation and relevant uncertainty. Identify which predicted conditions have confirmation observations and which are only calculations. Keep impossible or unapproved combinations outside the proposed experiment.

Before using an extended prediction to choose equipment or claim performance, ask what additional information would assess that application. Confirmation work should be feasible and connected to the intended decision. A model can guide a useful next question without being presented as a released specification for conditions never tested.

Trial discussion worksheet

This blank worksheet is for your own project. It contains no H M machine trial result.

Customer Questions

Primary References

These references support the technical principles discussed in this guide. The worked examples and review questions are educational.

• Primary reference: NIST Engineering Statistics Handbook