Verification

involves;

a. Sensitivity analysis: examining whether the model consistently reproduces patterns or if the results are sensitive to changes in the model's parameters. Two approaches commonly used are local sensitivity analysis (assessing the sensitivity of the model to changes in individual parameters and involves varying one parameter at a time while keeping the others fixed and observing the resulting changes in the model's outputs) and global sensitivity analysis (involves varying all parameters simultaneously).

b. Uncertainty analysis: focuses on quantifying the uncertainty associated with the model's outputs. It investigates how different plausible values of the model's parameters affect the reliability and variability of the results. 

c. Robustness analysis: explores the resilience of the model's results and conclusions to changes in its structure. It investigates whether the model's findings remain consistent and meaningful when the underlying assumptions, rules, or mechanisms are modified. 

Two methods to check stability (robustness):

• Plotting the acumulative average of the state variable (output) over an increasing number of runs. 
• The coefficient of variation is defined as the ratio between the standard deviation of a sample and the mean of that sample resulting in the following formula: Cv = бµ

To verify or understand the model's output, evaluating the simulation history is important. This can be done by;

a. analyzing key events chronologically,

b. examining the history of individual agents, and considering a global viewpoint for emergent patterns.

Method 1: Plotting Accumulative Average

• Concept: This method involves running the ABM multiple times (simulations) and calculating the average value of a chosen state variable (output) at each time step. As the number of simulations increases, the plot of the accumulated average should ideally converge towards a stable value.
• Process:
  1. Define the state variable of interest (e.g., average population size, number of infected individuals).
  2. Run the ABM multiple times (e.g., 100 simulations).
  3. For each time step, calculate the average value of the chosen state variable across all simulations.
  4. Plot the accumulated average of the state variable over the increasing number of time steps.
  Method 2: Coefficient of Variation (CV)

• Concept: The coefficient of variation (CV) is a measure of relative variability compared to the mean value. In the context of ABMs, it can be used to assess how much the model's output (state variable) fluctuates across multiple simulations.

• Calculation: CV = (Standard deviation of the state variable) / (Mean of the state variable)
• Interpretation:
  • A low CV (close to 0) indicates low variability, suggesting consistent model behavior across simulations.
  • A high CV indicates high variability, suggesting the model's output might be sensitive to initial conditions or random factors, potentially raising concerns about robustness.