階層線性模式HLM (hierarchical linear model) 是以層次分析方式來處理巢狀或內屬(nested structure) 資料, 並以隨機效果(random effect) 來估計截距項以及斜率項。因此, 本文以台灣地區23個縣(市) 地區之住宅調查統計資料作為分析樣本, 實證結果顯示住宅建物特徵與價格的關係, 會隨著縣(市) 地區不同而有所差異, 且區域特徵不僅對住宅價格有直接效果, 亦會在住宅建物特徵與住宅價格間產生調節效果。最後並與傳統迴歸模型分析結果作一比較, 其結果顯示, 傳統迴歸模型由於忽略了住宅空間效果易造成係數標準誤的低估, 造成顯著性考驗高估與型I 錯誤擴大的問題。

A hierarchical linear model (HLM) is used to process a nested structure by
hierarchical analysis and use random effects to estimate the intercept term and the slope term in a model. This study employs the housing statistics of 23 counties (cities) in Taiwan as the analysis sample to explore the impact of the characteristics of housing location and characteristics of housing structure on housing prices, and further to clarify the hierarchical structure of the analyzed data. The empirical results reveal that the relationship between characteristics of housing structure and housing prices significantly vary across different counties (cities), and characteristics of housing location not only have direct effect on housing prices but also have a mediating effect on the relationship between characteristics of housing structure and housing prices. Finally, the empirical results are compared with the results derived from a traditional regression models. It is found that since the traditional housing hedonic price model ignores the housing spatial effect, it may underestimate the standard error of coefficients, resulting in overestimation of the significance test and a larger Type I error.