8 - Analyzing Agent Based Models

Question 1

When do you think about measurements ?

Answer 1

• think of what you want to measure and analyze before you build your model
• should be clearly linked with research question or goal

Question 2

Why do we need multiple runs?

Answer 2

• even though it is a simulation we still have randomness in the sample
• Sampling error > we want to be confident that our results are robust

Question 3

What effects the number of runs?

Answer 3

• Number of parameters
• Convention
• Variance stability
• Power analysis (detecting effects)

Question 4

How can you determine the numbers of run you need?

Answer 4

• Variance stability
• Power analysis

Question 5

How do you conduct variance stability?

Answer 5

• Look at variance across sample sizes (runs)
• Set threshold
• take outcome measure - look at how much variance there is overtime
• set runnumber

Question 6

What are the prerequisites for power analysis?

Answer 6

Need to know what kind of analysis you plan to do and what size effect you might have (or a range)

Question 7

What is parameter sweeping/sensitivity analysis?

Answer 7

• systematically checking the plausible range of parameter values
• assess the importance of parameters based on model behaviour

Question 8

What are the types of data to analyze from ABMs?

Answer 8

• statistical
• graphical
• network-based
• spatial

Question 9

What is model validation?

Answer 9

Determine whether the model corresponds to the real-world phenomena

Question 10

What is model verification?

Answer 10

Determine whether implemented model corresponds to target conceptual model

Question 11

What is model replication?

Answer 11

Someone else implements the model and sees if the results are consistent