On the effective geographic sample size