Building valid, credible, and appropriately detailed simulation models - Statistical procedures for comparing real-world observations and simulation output data

8 important questions on Building valid, credible, and appropriately detailed simulation models - Statistical procedures for comparing real-world observations and simulation output data

When do we have nonstationary data:

If the distributions of the successive observations change over time.

What is a consequence of having autocorrelated data?

None of the classical tests is directly applicable.

What are 2 procedures for comparing real-world obsevations and simulation output data?

1. Basic inspection approach.
2. Correlated inspection approach.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

Correlated inspection approach:

Drive the model with historical input data, such that the system and the model experience the same observations from the input data, and compare outputs.

What is the danger of the basic inspection approach?

Each statistic is a sample of size 1 from some underlying population, making this idea vulnerable to randomness.

What are the steps in a CI-approach?

- m independent sets of data from the system (Xj i.i.d. with mean mu_x)
- n independent sets of data from the model (Yj i.i.d. with mean mu_y).
- Construct CI for mu_x - mu_x.
- If 0 is not in the CI, the observed difference is statistically significant at level alpha.
- If the magnitude of the difference is large enough to invalidate any inferences about the system that would be derived from the model, the difference is practically significant.

What are 3 time-series approaches?

- Spectral-analysis approach.
- Hsu & Hunter approach: fitting parametric time series model to each set of output data and applying a hypothesis test to see whether the 2 models appear to be the same.
- Chen & Sargent approach: CI for the difference between the steady-state mean of the system vs. model, based on Schruben's standardized time-series approach.

What are the requirements for a spectral-analysis approach?

- Output processes must to be covariance-stationary.
- High level of mathematical sophistication is needed.

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo