自Horst (1941)提出suppressor變數一詞起，由於後續許多研究採用不盡相同思維之著眼點，也就衍生出許多不同定義的suppressor變數。Horst (1941)著重在判定係數的變化，Darlington (1968)、Conger (1974) 及 Cohen and Cohen (1975)則著重在迴歸係數的變化， Velicer (1978)則改用semipartial correlation coefficient來定義suppressor變數，再度將焦點轉回Horst (1941)的思維。Currie and Korabinski (1984)引進enhancement一詞，以便與suppression有所區分。 為了釐清這些紛擾的名詞定義，第二章、三章中，我們分別回顧enhancement與suppression兩現象。第四章中，我們針對enhancement、suppression兩種現象的關聯性，依據四種不同的面向進行比較。第五章，我們針對suppressor變數存在的情況下，藉由模擬實驗的方式，探討stepwise regression、forward selection、backward elimination三種變數選取方式的可能缺憾。第六章為總結。 / Since Horst (1941) introduced the term of suppressor variable, many different definitions of suppressor variables have appeared in literature. Originally, Horst (1941) based the definition on the coefficient of determination. Darlington (1968), Conger (1974) and Cohen and Cohen (1975) paid more attention on the regression coefficients instead. On the other hand, Velicer (1978) used semipartial correlation coefficient to define a suppressor variable, and directed the focus back to that of Horst (1941). In order to differentiate the two similarly related ideas, Currie and Korabinski (1984) proposed the term of enhancement to describe exclusively the situations reflected by the definition of Horst (1941) or Velicer (1978). In order to clarify the ambiguities resulting from various definitions of suppressor variable in literature, we first reviewed enhancement and suppression respectively in Chapters 2 and 3. In Chapter 4, we investigated their relationships from four different perspectives. In Chapter 5, we studied the possible drawbacks on using stepwise regression, forward selection, and backward elimination these three variable selection procedures on the presence of a suppressor variable. Conclusions are provided in Chapter 6.