透過您的圖書館登入
IP:13.58.77.98
  • 期刊
  • OpenAccess

統計調查中遺漏值處理之研究-以臺灣地區工程技術顧問服務業調查為例

Imputation Method for Missing Data in Survey-Engineering Technology Consultancy Service Survey in Taiwan Area

摘要


在求真求善的統計學理的探討上,如何使統計調查結果更接近母體的態樣,一直是努力研究的方向,故處理遺漏值以減少調查的非抽樣誤差,仍有一些意義與科學研究之價值。準此,本文針對統計調查中遺漏值發生的原因以及處理方法加以彙整研究,並以行政院公共工程委員會所辦理之「臺灣地區工程技術顧問服務業調查」為例,以SOLAS模擬執行熱卡法及多重插補並評估其效果。在模擬過程中,首先找出問卷中基本問項與插補問項相關較大者後,以隨機遺漏之方式產生不同遺漏比率之資料集,分別執行SOLAS之熱卡法、多重插補法並評估模擬結果。本文所探討遺漏值發生的原因,除可供調查過程中事前預防措施的參考,亦可作為選擇處理遺漏值方法的重要評估因素。在實例研究中,多重插補模擬結果之平均數與原始得點數的差異小,顯示效果良好,並可快速產生完整的資料集,供研究者做深入的分析;此外,遺漏值的比率與各種插補法的效果有關,可供本調查選擇插補法之參考,此亦為本研究的重要發現。

關鍵字

遺漏值 多重插補 熱卡法

並列摘要


In discussion the statistics study, how is it count investigation result close to pattern of parent to make, it has been a direction made great efforts to study all the time. It was not the sampling error that dealt with the omitting value in order to reduce the investigation, there is value of some meanings and scientific research. This text gathers together and exactly studies to the missing value reason and method that deals with in the census, and 'Technical Adviser of Taiwan Project Service Trade Investigate' as the example with executive organ public project what committee handle, carry out Hot-Deck and multiple imputation and assess its result again with SOLAS simulation. In the course of imitating, find out and ask basically in the questionnaire that one is with inserts and mends asking that one is relevant after the bigger one at first, different materials collections which omits the rate produce in way that is omitted at random, carry out the Hot-Deck of SOLAS, multiple imputation and assessment. The reason to discussion the reason of missing value, divided by is suitable for the reference of the precautionary measures in advance in the investigation course, can also deal with the important assessment factor of the omitting value method as the choice. Among case study, many to is it mend average, simulation of result with so primitive that the difference of counting little to insert again, show that the result is good, and can produce the intact materials collection fast, for the researcher to make deep analysis; In addition, missing value rate and various kinds of results of inserting the reinforcing method are relevant, for this investigation to choose to insert the reference of the reinforcing method, this is also an important discovery of this research.

並列關鍵字

missing value multiple imputation Hot-Deck SOLAS

參考文獻


翁彰佑、程爾觀(1991)。隨機遺失資料插補法估計效用之比較。中國統計學報。29(2)
劉長萱、蔡政豐(1996)。問卷調查的不完整取樣設計。調查研究。3
陳信木、林佳瑩(1997)。調查資料之遺漏值的處置—以熱卡插補法爲例。調查研究。3
金勇進(1996)。非抽樣誤差分析。中國統計出版社。
黃登源(1998)。應用迴歸分析。華泰文化事業公司。

被引用紀錄


程于庭(2013)。遺失資料插補法在最適資產配置投資組合上之應用與比較-以臺灣證券市場為例〔碩士論文,淡江大學〕。華藝線上圖書館。https://doi.org/10.6846/TKU.2013.00602
姜憶如(2009)。基隆市國中學生對法定性傳染病之認識及其相關因素初探〔碩士論文,中山醫學大學〕。華藝線上圖書館。https://doi.org/10.6834/CSMU.2009.00080
洪靜茹(2013)。缺失資料處理對多變量變異數分析(MANOVA)的影響〔碩士論文,國立臺北大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0023-0602201318131300

延伸閱讀