寻求更便捷的Agent种群统计方法：覆盖当前及已移除个体

Great question—dealing with historical agent stats can feel clunky when relying solely on the built-in Statistics panel or SD flows (like the QALY calculation in the Basic Health Economics model). Here are a few cleaner, more efficient approaches I’ve used to track both active and removed agents without the overhead of per-stat SD flows:

1. Global Accumulator Variables + Agent Local Tracking #

This is my go-to for straightforward, low-overhead stats:

• Create global variables (e.g., totalAgentsEverCreated, sumRemovedAgentQALYs) to track cumulative values from removed agents.
• In your agent’s code:
  • When an agent is initialized, increment totalAgentsEverCreated by 1.
  • Maintain a local variable (e.g., lifetimeQALY) that updates each time step with the agent’s current contribution (Quality of Life * time step duration).
  • Right before the agent is removed from the model, add its lifetimeQALY to sumRemovedAgentQALYs.
• To get the total for all agents (active + removed), just combine the global accumulator with the live statistic from the Population panel:

  totalAllQALYs = sumRemovedAgentQALYs + population.statistics.sum(lifetimeQALY)
  

This avoids the need for SD flows entirely and keeps your model logic focused on agent behavior rather than flow wiring.

2. Database Table for Historical Agent Records #

For models where you need flexible, queryable historical data (e.g., breaking down stats by agent group, time period, or other attributes):

• Set up a simple database table (most ABM platforms support this natively) with columns like agentId, birthTime, removalTime, totalQALY, healthStatusHistory, etc.
• Insert a row into the table when an agent is created, and update the removalTime and final stats (like totalQALY) when the agent is removed.
• You can then run direct queries on the table to calculate any historical statistic you need—for example, summing totalQALY across all rows gives you the full population’s lifetime QALYs, no need to split active/removed manually.

3. "Archived Agents" List (For Flexible Attribute Access) #

If you need occasional access to full agent attributes (not just cumulative stats), maintain a global list to store references to all agents that have ever existed:

• Create a global List<Person> (or your agent type) called allAgentsEver.
• Add each new agent to this list when it’s initialized.
• Even when agents are removed from the model’s main population, they’ll remain in this list (just be mindful of memory usage for large agent populations—you can mitigate this by storing lightweight data objects instead of full agent instances if needed).
• To calculate stats, you can directly iterate over the list:

  double totalQALYs = allAgentsEver.stream()
      .mapToDouble(agent -> agent.lifetimeQALY)
      .sum();
  

This is great for ad-hoc analysis or when you need to compare active and removed agents on multiple attributes.

Comparing to the Basic Health Economics Model #

The SD flow approach in that model works for integrating QALYs over time, but it’s overkill if you just need to track cumulative stats across all agents. The methods above eliminate the need for flow diagrams, reduce model clutter, and make it easier to add new historical stats without reworking your SD structure.

内容的提问来源于stack exchange，提问作者mivandev