請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51426
標題: | 時間序列計數資料之統計分析:以台灣地區抗蛇毒血清使用量為探討 Statistical Analysis of Time Series Count Data: A Study of Consumption of Snake Antivenom in Taiwan |
作者: | Chen-Ku Wen 溫振谷 |
指導教授: | 張淑惠(Shu-Hui Chang) |
關鍵字: | 抗蛇毒血清,氣象因素,時間序列,自迴歸,自迴歸整合移動平均, snake antivenom,meteorological factors,time series,autoregressive,autoregressive integrated moving average, |
出版年 : | 2016 |
學位: | 碩士 |
摘要: | 研究背景
台灣每年約有1000人因蛇咬傷至醫療院所求診,及時注射正確種類的抗蛇毒血清對降低重症及死亡率具有關鍵的影響。目前使用的4種抗蛇毒血清中,以抗龜殼花及赤尾鮐蛇毒血清凍晶注射劑(批號代碼:Freeze Hemorrhagic ,以下稱FH) 每年使用約2600瓶為最多,但該種藥品目前年供應量約3000餘瓶,僅能達到人用供需平衡。毒蛇咬傷事件有明顯的季節效應,顯示抗蛇毒血清使用量亦隨氣象因素而變化,然而目前並無文獻探討氣象因素與抗蛇毒血清使用量的關係。 研究目的 藉由分析全國性的資料庫,本論文研究目的為: (1) 了解氣象因素對FH使用量的影響, (2) 評估以不同統計模式配適FH使用量的差異,探討應用於預測的可能性。 研究方法 本研究利用2008-2012的健保資料庫及中央氣象局的氣象資料,以台灣地區抗蛇毒血清使用量最多的FH做為分析種類,並選擇FH使用量前2名的縣市作為研究區域,分析氣象因素與FH使用量之間的簡單相關性,包含斯皮爾曼等級相關、淨相關及自相關。以不同模式配適度標準來比較以下統計模式,包含類卜瓦松迴歸模式、負二項迴歸模式、自迴歸殘差之線性迴歸模式與及自迴歸整合移動平均模式等四種統計迴歸模式。 研究結果 2008-2012年FH使用量最多的前2名縣市分別為台中市及花蓮縣。多變數分析結果顯示台中市以平均濕度及平均溫度為顯著因素,花蓮縣則以降雨量及前2個月FH使用量為顯著因素。台中市在調整平均濕度後,當月平均溫度高於33.6℃或低於23.6℃時,該月平均的FH使用量會減少;花蓮縣在調整前2個月FH使用量後,該月平均的FH使用量會隨降雨量升高而增加。以一般線性迴歸配適開根號處理後的FH使用量,其AIC 及BIC較其他模式為小。 結論 本研究結果發現不同縣市受不同氣象因素影響抗蛇毒血清FH的使用量,氣象因素及過去FH使用量是建立預測模型時需考慮的重要變數。 Background There are about 1,000 snakebite victims who seek medical treatment in Taiwan every year. Injecting correct snake antivenom timely plays a key role in reducing morbidity and mortality. Among the four types of antivenom currently used, the consumption of antivenin of TR. mucrosquamatus and TR. gramineus(batch code: Freeze Hemorrhagic, abbreviation FH) ,consumed 2,600 doses annually, is the most. Nevertheless, the supply of FH approximately 3,000 doses per year, only enough for human needs. Snakebite incidents show a manifest seasonal pattern. It indicates that the consumption of snake antivenom might change along with the meteorological factors. However, to our best knowledge, we do not find any published literature so far in studying the association between meteorological factors and the consumption of snake antivenom. Aims By using nationwide databases, the objectives of my thesis are (1) to clarify the effects of meteorological factors on the consumption of FH; (2) to evaluate the goodness of fit of different statistical models and to discuss the potential in prediction. Material and Methods This research focuses on the analysis of the FH massive consumption from 2008-2012 National Health Insurance Research Database together with the meteorological data from Central Weather Bureau. Two counties or cities which used FH most during 2008-2012 are selected as the study areas. The correlation between meteorological factors and consumption of FH were analyzed by spearman rank correlation, partial correlation, and autocorrelation. Various goodness-of-fit criteria were used to compare different statistical models, including quasi-Poisson regression model, negative binomial regression model, autoregressive error model, and autoregressive integrated moving average model. Results The two areas consuming FH most are Taichung city and Hualien county between 2008-2012. The results of multiple regression model show that monthly mean humidity and monthly mean temperatures were significant factors in Taichung, and the results in Hualien show that precipitation, first and second-order autocorrelation of consumption of FH were significant factors. After controlling monthly mean humidity, the average monthly consumption of FH in Taichung tend to be lower when monthly mean temperature is higher than 33.6℃ or lower than 23.6℃. The average monthly consumption of FH in Hualien tends to be higher as precipitation increases after controlling first and second-order autocorrelation of consumption of FH. Based on Akaike information criterion and Bayesian information criterion, the ordinary linear regression for the square root of the monthly FH consumption tends to be a better prediction model than other models. Conclusion This research discovered that consumption of FH is affected by different meteorological factors in different cities. Climate factors and the former consumption of FH may be important in building prediction model. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51426 |
全文授權: | 有償授權 |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 2.79 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。