- 2074 Magh · 5 marks
Compare verification, validation, calibration
Short answer
Verification ensures the program matches the conceptual model (e.g., checking shortest-queue logic). Validation confirms the model accurately represents reality (e.g., statistically comparing simulated and real waiting times). Calibration iteratively adjusts parameters until output matches observed data (e.g., changing service time from 2 to 2.8 min).
Comparisons between verification, validation and calibration of a model using example.
Verification checks that the model is built correctly (the computer program matches the conceptual model). Validation checks that the right model has been built (it is an accurate representation of the real system). Calibration is the iterative process of comparing the model with the real system and adjusting the model until agreement is acceptable.
Comparison
| Point | Verification | Validation | Calibration |
|---|---|---|---|
| Question asked | Is the model built right? | Is it the right model? | How can the model be adjusted to match? |
| Compares | Conceptual model vs program | Model vs real system | Model output vs real data |
| Concerned with | Programming, logic, debugging | Accuracy, assumptions, credibility | Parameter and structure tuning |
| Methods | Walk-throughs, trace, debugger, simple test cases | Face validity, assumption tests, input-output comparison | Repeated comparison and modification |
| Result | Error-free implementation | Confidence that model is usable | Better agreement with reality |
| When done | During and after coding | After a working model exists | In loop with validation |
Relationship
Real system -> Conceptual model -> Operational model
validation <---------------------->
verification (conceptual to program)
Compare model with system -> adjust -> compare ... (calibration)
Calibration and validation go together: compare, find a difference, modify the model, and compare again until the model is accurate enough for its purpose.
Example: bank with three tellers
- Verification: the conceptual model says a customer joins the shortest queue. The programmer prints a trace of 20 arrivals and checks that each customer really joined the shortest queue, that clock times increase, and that the number of departures never exceeds arrivals. A bug such as the clock not advancing is found and fixed.
- Validation: the model is run with the same arrival pattern recorded on Monday. The model gives average waiting time 3.8 min; the observed value is 4.0 min. A statistical test shows no significant difference, so the model is accepted as valid for this purpose.
- Calibration: suppose the model gave 2.5 min against 4.0 min observed. The analyst finds that service time was assumed 2 min but includes no time for paperwork. Service time is changed to 2.8 min, and the model is rerun until the waiting time matches the real value within tolerance.
✓ Answer checked by Claude Opus 5.5 (high effort)
Related questions
- Model building, verification, validation steps · Simulation and Modeling, chapter 7
- Calibration and validation of models · Simulation and Modeling, chapter 7 · asked 2×
- Iterative process of model calibration · Simulation and Modeling, chapter 7 · asked 4×
- Verification versus validation; critical systems · Software Engineering, chapter 7
- Compare software models; propose a fourth model · Instrumentation II, chapter 9