공간통계기법을 이용한 암 발생율과 지리·환경적 특성과의 연관성 분석 = Analysis of association between cancer incidence and geography·environmental characteristics using spatial statistical methods

다국어 초록 (Multilingual Abstract)

Cancer is a disease that ranks first as a major cause of death for Korean as of 2017, and it is known to be mainly caused by environmental factors. Despite the many deaths due to cancer, there is a lack of research how environmental factors are related with regional causes of cancer in Korea. The purpose of this study was to investigate how different the incidence of various cancer including thyroid cancer, colorectal cancer, gastric cancer, lung cancer, liver cancer, prostate cancer, and breast cancer from geographical and environmental factors. For this, the incidence of cancer by administrative districts (si-gun-gu level) was selected as a dependent variable and 13 variables that are expected to affect the causing of cancer are selected as the independent variables.
Global Moran's I and Local Moran's I statistics were used to order to identify spatial dependency in dependent variables. OLS and spatial regression analyses were utilzed to identify the factors affecting cancer incidence and to consider spatial dependence in regression models. As a result of checking Moran index value and the LISA Cluster map, spatial autocorrelation was confirmed in the pattern of cancer incidence. After that, through OLS regression analysis, I figured out which variables had affect the cancer development without assuming spatial autocorrelation. Based on the results, spatial regression analysis was performed to confirm that there is a spatial dependence between the onset of cancer and major risk factors. The results of the analysis show that the spatial error model is the most suitable for all cancer. The number of automobile registration per household, the ratio of population over college, the ratio of buildings before 1980, and the ratio of buildings before 1995 are significant variables in positive direction. In the case of thyroid cancer, the spatial lag model is the most suitable. Only the current smoking rate is significant in the negative direction. Spatial error model is most suitable for colorectal cancer. The number of automobile registration per household, the ratio of buildings before 1995, the population density, and the current smoking rate were significant in positive direction. On the other land, the population per doctor was significant in the negative direction. In the case of stomach cancer, the spatial lag model is the most suitable. while the ratio of manufacturing workers and the ratio of buildings before 1995 were significant in positive direction, the Population per doctor was significant in the negative direction. In case of lung cancer, the spatial error model is most suitable. The sex ratio, the ratio of buildings before 1995, and the current smoking rate were significant in positive direction. But, the ratio of apartment was significant in the negative direction. In the case of liver cancer, the spatial lag model is suitable. The ratio of buildings before 1995 and the high risk drinking rate were significant in positive direction, but the ratio of population over college and the ratio of green field were significant in the negative direction. In the prostate cancer, the spatial lag model is the most suitable. The sex ratio and the ratio of population over college were significant in the positive direction, and the ratio of apartment and the ratio of manufacturing workers were significant in the negative direction. In breast cancer, the spatial error model is most suitable, and the population density, number of automobile registration per household, the ratio of the population over college, and the current smoking rate were significant in the positive direction. But the sex ratio and the ratio of apartment were significant in the negative direction.
This study will provide useful information on how to control the geographical environmental risk factors at the regional level in order to overcome and prevent each cancer based on the analysis of geographical environment factors affecting cancer development and cancer.

목차 (Table of Contents)

• Ⅰ. 서론 1
• 1. 연구의 배경 및 목적 1
• 2. 선행 연구 2
• 3. 연구 내용 구성 4
• Ⅱ. 변수 선정 및 분석 방법 6
• 1. 변수 선정 및 정의 6
• 1) 종속변수 7
• 2) 독립변수 7
• 3) 변수 선정의 한계 9
• 4) 기술통계 10
• 5) 상관분석 12
• 2. 분석 방법 14
• 1) 암의 공간적 자기상관 14
• 2) 공간회귀모형 16
• Ⅲ. 암 발생에 영향을 주는 지리환경적 요인 19
• 1. 암 발생의 공간적 자기상관 19
• 1) 전역적 Moran’s I 통계량 19
• 2) local Moran’s I 통계량 19
• 2. 암 발생에 영향을 주는 요인 분석: OLS 접근 28
• 3. 암 발생에 영향을 주는 요인 분석: 공간회귀 접근 49
• 1) 잔차의 Moran’s I 통계량 49
• 2) 모델 간 비교 분석 50
• Ⅳ. 결론 69
• 참고 문헌 71