2010-08-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/647155摘要：自1990年來針對地區對健康的影響(health and place)研究文獻己累積甚多，己逐漸運用多階層迴歸模型，以便同時考量個人層次的社會經濟位置與居住地區的社經結構或財富不平等對個人健康的影響，這種研究取徑能夠避免過去研究常以地區為分析單位所引發的「地區繆誤」(ecological fallacy)，然而卻無法考慮地區間互動的動態關係。因為在這些研究中，地區會依照慣例被定義為人口普查區或行政部門，這種定義經常會專注於內部社區的特點，而假定地區間沒有互動，亦即，單一地區的影響因子不會有外溢效果（spillover effects）。根據Morenoff et al. (2001), Morenoff (2003)以及Caughy et al. (2007)所提出的研究，社會過程會擴大且超越了地理界線，相同的健康風險會擴散至對鄰近地區。然而，如何理解或在分析架構中納入空間關係結構 (空間效果)及空間異質性對健康的影響，則是一個相當重要的課題。這些過去地區與健康(health and place)的研究文獻，多運用近年來較為流行的多階層迴歸模型，以便同時考量個人層次的社會經濟位置與居住地區的社經結構或財富不平等的程度，這種研究取徑能夠避免過去研究常以地區為分析單位所引發的「地區繆誤」(ecological fallacy)的問題(Black & Macinko, 2008)，然而卻無法考慮地區間互動的動態關係。因為在這些研究中，地區會依照慣例被定義為人口普查區或行政部門，這種定義經常會專注於內部社區的特點，而假定地區間沒有互動，亦即，單一地區的影響因子不會有外溢效果(spillover effects)。根據Morenoff et al. (2001), Morenoff (2003)以及Caughy etal. (2007)所提出的研究，社會過程會擴大且超越了地理界線，對鄰近地區也會形成相同的健康風險。因此，影響肥胖風險的因素除了出現在單一社區裡面，也可能透過地區間相互影響、人群互動，而逐漸擴散至其鄰近的地地區。然而，如何理解或在分析架構中納入空間關係結構 (空間效果)及空間異質性對健康的影響，是一個相當重要的研究課題。本研究旨以橫斷面資料及空間迴歸(spatial lag model, spatial error model)，地理加權迴歸(Geographically weighted regression, GWR)分析模型，估計空間聚集模式及收斂後的空間外溢效果(第一年計劃)，以及同時考量空間關係結構及社會互動關係網絡(例如經濟交易網絡或人口遷移網絡)，可進一步將空間迴歸模型擴展成「雙矩陣空間迴歸模型」以同時表現兩不同空間矩陣的相依關係，並同時估計兩空間矩陣的參數(地理鄰近空間矩陣及社經活動聯繫矩陣)對健康及疾病的影響(第二年計劃)，並在時間、空間皆會改變的歷史架構下，運用Franzese, et al. (2006)所提出的空間Panel 模型，以及多階層模型中的成長模型(Growth model)，再整合Snijders (2001; 2005)所提出動態貫時序社會網絡分析，重新建立一個新的健康區域不平等研究典範，架構「人、時、地三向度整合」的健康不平等研究取徑(第三年計劃)。此一研究取徑旨在納入空間關係結構的影響，以估計個人層次及地區層次的健康影響因子。這是一個跨學科整合的研究，對未來的研究有一定的典範作用。<br> Abstract: Research on estimating neighborhood effect on health inequality has accumulated since1990. Multilevel models are often used as the standard statistical methods to account for therelative effect of neighborhood and individual SES on health. However, the impact ofneighborhood effects may be either within or among neighborhoods. Previous studies ofneighborhood effects on health have modeled neighborhoods as if they existed independentlyand failed to consider interrelations with nearby neighborhoods. In these studies,neighborhoods were conventionally defined as census tracts or administrative units; suchdefinitions typically focused on the internal characteristics of neighborhoods and ignored anyinfluences on health that may result from the interactions within a broad socio-economiccontext. Neighborhoods were also assumed to have no interaction; that is, neighborhoods hadno spillover effects. Thus, the potential embeddedness of neighborhoods within a larger socialenvironment was generally overlooked.The study is aimed to address the important theoretical question that whether theneighborhood effects on health should not be treated as islands unto themselves, becausefactors affecting obesity risk in one place are also likely to affect obesity risk in nearby places.This concept of spatially dependent neighborhoods has significant methodological implicationswhen interpreting neighborhood effects. The pattern of spatial externality should be alsoconsidered in the study of neighborhood effect on health. The purposes of this study are:1) byemploying an integrated approach incorporating multilevel models and spatial regressionmodels to address the issue of spatial dependence in the neighborhood effect on health; 2) byemploying the dual weight matrixes in spatial regression model to address relative importanceof social networks and spatial weight matrix on health outcomes;3) by employing space-timeanalytical modules to address the issue of how the changing neighborhood structures affectindividual-level health outcomes.This study is an interdisciplinary effort to build up an analytical framework to addressthe issue of how people, place and time interact with each other and generate long-termeffects on individual-level health outcomes. We also address the long-term changes in thepolitical-social-spatial structure in Taiwan. Methodologically, geographical informationsystem (GIS), spatial econometrics, and multilevel models are integrated to build up ananalytical framework to address the issues.