從研究方法的觀點，欲探討某一學術領域之論文寫作的研究趨勢，倘若不考慮時間因素，祇是單純合併不同時間點之資料進行分析，會拘泥於橫斷面的觀察，僅能利用靜態比較方式反映出研究主題與方法選擇的差異，恐無法檢證時間遞移對論文研究趨勢變化的具體效應；另一方面，若祇單純考慮時間序列資料的分析方法，仍未將環境系絡的結構因素納入考量，則又無可避免地會喪失攸關縱橫時空之資料（panel data）的完整訊息。因之，本研究嘗試以台灣公共行政領域的碩士論文為分析標的，提出一個有待檢驗的基本命題：雖然影響各校相關公共行政系所碩士論文總產量，如招生人數或畢業生多，論文數量自然就會增加，但是，論文產量研究趨勢的歷年變化會依研究主題、研究方法、研究對象而有所不同；尚且，由於各校發展時空環境不同，如學校所在區域、公行系所歷史、指導教授背景等屬結構性總體脈絡變數仍存有交互影響的效果。 本研究利用多層次分析（multilevel analysis）的方法，檢證出各校公共行政碩士論文產量的增加，並不全然是隨著時間而等比例增加，而是依硏究主題、資料分析方法、硏究對象而有不同比例的增加趨勢，反映了議題與方法選擇的消長變化；且伴隨著結構性脈絡變數的變化也可能會影響論文產出，而非單純取決於研究生個人的興趣。經由此篇應用多層次分析方法的引介，不但可對台灣公共行政學領域相關校系的長期研究走向的變異來源和意涵提供更深刻的理解，也可確認總體層次與個體層次的變數對論文產量的影響大小及方向。

From a methodological point of view, when examining the overallresearch output and trends of a growing academic discipline, the effects of timeshould be taken into consideration, for otherwise analysis is reduced to crosssectional,comparative static analysis of differences in research topics andmethods. On the other hand, if time effects were the only variables taken intoaccount, analysis would also, inevitably, squander vast information that iscontained in various contextual variables affecting research production. Thefundamental proposition of this paper is that the choice of research topic andmethod is strongly influenced by contextual and other environmental factors.Using graduate theses in Taiwan public administration as data, this study seeksto investigate not only the dynamic interaction of time-environment effects, butalso those contextual factors that shape Taiwanese students’ selection ofresearch topics and methods. Our research includes both micro-level variables (i.e., research topic and method) and macro-level variables, such as background of students’ thesisadvisors, location and historical background of graduate programs andinstitutes. By taking a multilevel perspective, this study demonstrates that thenumber of graduate theses in public administration does not necessarilyincrease proportionally over time; rather it is subject to the growth and declineof research topics and methods. Moreover, thesis output is also influenced bystructural changes in the contextual variables, and does not entirely hinge onstudents’ personal research interests. We conclude that the introduction ofmulti-level analysis does not only provide a deeper understanding of thesources of variation in long-term research trends, but also can determine theimpact –both size and direction– of macro- and micro-level variables ongraduate thesis production in Taiwan’s public administration.