V&V are two distinct but essential steps for ensuring the credibility of a simulation model. An analyst’s key challenge is to validate the model, as it is only useful if it is an accurate representation of the actual system.

• Verification: The process of ensuring the model’s implementation and its data align with the developer’s conceptual description and specifications.
• Validation: The process of comparing the conceptual model’s representation to the real system.

V&V Techniques

Verification Techniques:

• Writing and debugging the program in sub-programs using skilled programming.
• Employing a “Structured Walk-through” policy where multiple people review the program.
• Tracing intermediate results and comparing them with observed outcomes.
• Checking the model’s output against a variety of input combinations.
• Comparing the final simulation result with known analytic results.

Validation Techniques:

1. Design for High Validity:
   • Discuss the model design with system experts.
   • Maintain interaction with the client throughout the development process.
   • Have system experts supervise the output.
2. Test Model Assumptions:
   • Quantitatively test the model by applying assumption data.
   • Perform sensitivity analysis to observe how results change when significant changes are made to input data.
3. Determine Representative Output:
   • Assess how closely the simulation output matches the real system’s output.
   • Use a Turing Test, where data is presented in the system format for experts to evaluate.
   • Employ statistical methods (e.g., chi-square test, Kolmogorov-Smirnov test) to compare model output with real system data.

Data Comparison Approaches

• Validating an Existing System: Use real-world inputs in the model and compare its output with the real system’s output. This may involve statistical hypothesis testing for metrics like waiting time or idle time.
• Validating a New/Hypothetical System: When historical data is unavailable, validation relies on:
  • Subsystem Validity: Testing known subsystems within the larger model separately.
  • Internal Validity: Ensuring the model does not have a high degree of internal variance that would obscure the impact of input changes.
  • Sensitivity Analysis: Identifying sensitive parameters that require higher attention.
  • Face Validity: Rejecting a model if it operates on opposite logic, even if its behavior mimics the real system.