本文之目的在利用模型假設錯誤時，Bartlett第二等式的違背，提出一個變異數函數的推論方法，而重點則在變異數函數中過離散係數的部分。我們並使用拔靴法來估計過離散係數估計量的變異數。The Bartlett’s second identity collapses when model assumption is mistaken. We adopt the Poisson distribution as the working model and employ the breakdown of the Bartlett’s second identity to make inference about the overdispersion coefficients of count responses.
