具群組因子效果的條件邏輯模型探討

Abstract

條件邏輯模型是在1973年由McFadden所提出，它的基本架構和我們一般所熟悉的多項邏輯模型大體上是相對等的，雖然較少為人所知，但近年來也逐漸受到經濟領域方面研究的重視。不過如果想要利用這一類模型來分析具有某些群組結構的資料時，常常會有低估標準差的現象產生。在此計劃中我們藉由引入隨機效用參數來處理兩類不盡相似的群組結構所衍生出來的問題。此外我們也進一步就有序反應變數的相關議題，有所探討。

Proposed by McFadden(1973), the conditional logit model, an equivalent form of the well-known multinomial logit model for nominal outcomes, has gained its popularity in economic research of late. The model, however, appears to underestimate the standard error when certain group structure exists in a data set. By introducing random effect parameters into the model, we can easily cope with the problem associated with two types of group structure. In addition, we also investigate the estimation problems related to the generalized linear mixed models with ordinal responses.