目前台灣MBA(Master of Business Administration)行銷研究相關碩士論文大多著重於實證性研究,假如研究者未對行銷研究程序中每一個階段可能產生的錯誤予以切實的掌握,將導致提出的研究結果無法顯現正確之情況。故本研究透過近百篇2004至2006年台灣MBA行銷研究相關碩士論文,探討台灣MBA行銷研究相關碩士論文於抽樣調查與統計分析這兩大主題的應用情形。並且利用固定數額追蹤樣本數估計理論推導而得的抽樣誤差,以及假設檢定問題中的無異區間與操作特性函數方法,經由模擬與實證分析,瞭解台灣MBA行銷研究相關碩士論文的抽樣設計準確度與統計方法正確性,進而解釋分析結果所代表的意義以及提供建議。本研究結果發現在正規抽樣設計情況下,研究者藉由樣本估計母體特徵值,估計錯誤風險與有效樣本回收率之大小會影響估計結果的準確度。此外抽樣調查對最後未回收的樣本數採取部分追蹤時,抽樣調查結果會較接近研究者所要求的估計準確度。至於當研究者無處理未回收樣本,有效樣本數增加,或是實際參數與檢定假設的已知常數之間沒有太大差距時,統計方法的可靠性會隨之提升。最後藉由有處理未回收樣本之研究,可以發現採用固定數額追蹤未回收問卷者的資訊,在所考慮的已知常數值下正確地接受檢定假設的機率,會比尚未採用固定數額追蹤未回收問卷時的機率高,再次證明研究者有無處理未回收問卷,在行銷研究工作中佔有十分重要的地位。
Nowadays most MBA (Master of Business Administration) theses on marketing subject in Taiwan focus on empirical study. If researchers cannot manage the mistakes that are made possible at different stages of marketing research process, the correctness of the outcome of the researches may be doubted. This study carried out an analysis and a discussion of approximately 100 MBA marketing theses published from 2004 to 2006 in Taiwan in the practical application areas of sampling design and statistical method. The methods of study are indifference region, operating characteristic function, and sampling error conducted from a fixed amount of tracing sample size. Using these methods to simulate and examine the accuracy of sampling design and statistical analysis of MBA marketing theses in Taiwan, this study draws some conclusions and provides suggestions. This study concludes that under formal sampling design, the variety of statistical error and survey response rate will influence the accuracy of the estimation. The outcome of sampling survey will be close to the accuracy of the estimation as researchers expect, when researchers draw a fixed amount of tracing sample size by collecting the data from those who did not return their questionnaires. For researchers do not deal with incompletely sample data, the higher reliability of statistical analysis will be acquired if the smaller gap is between population parameter and constant given by null hypothesis, and the more effective sample sizes are used. This study also finds that researchers who deal with incompletely sample data by drawing a fixed amount of tracing sample size will obtain higher probabilities that correctly accept null hypothesis than others who do not deal with incompletely sample data. This leads to the conclusion that whether unreturned questionnaires are dealt with or not is an essential process in a marketing research.