Table 1.

Summary of Simulation Steps

Part I: Gathering Information Prior to the Simulation

Step 1: Calculate degree distribution and differential degree from the sampled data.

Step 2: Calculate ego network configuration distribution from the sampled data.

Part II: Setting up the Simulation

Step 3: Simulate network of size N with the degree distribution from Step 1; assign demographic characteristics to the nodes in the network (based on the sampled data).

Set ERG Model to Simulate Network From:

Step 4: Specify terms in the model.

Model terms capture:

Differential degree (nodecovariate term)

Homophily (absolute difference or mixing matrix)

Ego network configuration distribution (GWESP or alternative clustering term)

Step 5: Set initial coefficients on terms from Step 4.

Step 6: Constrain model on the observed degree distribution (from Step 1)

Part III: Simulation Procedure

Step 7: Simulate network using the model specified in Steps 4–6. Start from network simulated in Step 3.

Step 8: Compare homophily in simulated network (from Step 7) to homophily in sampled data. Update homophily coefficients if bias is found.

Step 9: Simulate new network using the updated coefficients from Step 8. Start from the network in Step 7.

Step 10: Use chi square value to compare ego network configuration distribution in simulated network (from Step 9) to ego network configuration distribution in sampled data (from Step 2)

Step 11: Update coefficient on clustering term to find better fitting network. A “better” network has a lower chi square value (compared to the chi square value from Step 10), or has an ego network configuration distribution closer to the empirical distribution. Steps 7–10 are repeated for each proposed change to the clustering term coefficient (with the new clustering term coefficient used in the set of coefficients).

Repeat Step 11 until the expected chi square value does not improve over the last iteration.