Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/37128
標題: 紅火蟻餌劑型殺蟲劑試驗數據的統計分析-一個縱向性數據的個案研究
Statistical Analysis of An Experiment to Evaluate Bait-Formulated Insecticides in Controlling Red Imported Fire Ant-A Case Study of Longitudinal Data Analysis
作者: 吳啟綸
Wu, Chi-Lun
關鍵字: red imported fire ant
入侵紅火蟻
repeated measurement
longitudinal data
重複量測
縱向性數據
出版社: 農藝學系
摘要: 台灣地區在二0 0 0年左右遭受入侵紅火蟻(Red imported fire ant)的入侵,在受感染的地區不僅會在醫藥、經濟上造成損失,也會對生態造成嚴重的危害。是故為了針對政府所核准的三種餌劑型殺蟲劑-芬普尼、百利普芬及賜諾殺進行藥效的評估,因此行政院所屬的桃園區農業改良場及台南區農業改良場分別於受感染的桃園與嘉義兩地進行試驗。在兩試驗地點,皆設有三個藥劑處理及一個對照處理,每處理均重複配置於三個10×10平方公尺的試區,在藥劑施用後八週內重複地量測每試區的活動蟻塚數。對於這些縱向性數據而言,我們可以探討以下四個問題:一、這些藥劑在防治入侵紅火蟻上是否有效?二、這三種藥劑在防治上是否有同樣的效率?三、藥劑處理前的活動蟻塚密度是否會對藥劑的效果造成影響?四、處理之間的活動蟻塚密度是否有不同的時間趨勢變化?基於上述幾點問題本文試以SAS/STAT所提供的不同程序予以分析,並比較不同程序的適當性。 首先,以各試區在試驗開始時的蟻塚數為基準,計算各試區在試驗開始後不同時點上的相對蟻塚密度,將相對蟻塚密度經對數轉換後與時間作圖。從圖中可以發現三種藥劑處理均呈現出明顯的防治效果,其下降的時間趨勢概呈二次函數。 其次,使用Proc GLM執行裂區設計的時間分析,從分析結果中發現百利普芬餌劑在防治入侵紅火蟻的效果上較比芬普尼餌劑和賜諾殺餌劑為差。然而因桃園地區的試驗數據違背Huynh-Feldt的假設,因此以裂區設計的變方分析模式所分析的結果並不可靠,而在Proc GLM中增加REPEATED指令事實上有相同的分析結果。 再者,將每一試區蟻塚的對數相對密度分別配適時間的二次函數,視初始蟻塚密度為併變因子,以變積分析的技術來探討試驗開始時各試區之初始蟻塚密度及不同藥劑處理對各試區蟻塚對數相對密度之時間變化趨勢的影響。從結果中可以發現,僅於桃園地區的芬普尼餌劑在防治入侵紅火蟻上有顯著的效應。然而因其違背變積分析的前提假設,所以在分析上的結果並不可靠。 最後,使用Proc MIXED執行一般線性混合模式,結果顯示分別於兩試驗地點的三種餌劑在防治入侵紅火蟻上皆有顯著的效果。就桃園地區的試驗而言,處理前的蟻塚密度對於藥劑的防治率並沒有顯著的影響,在時間效應上,芬普尼餌劑和賜諾殺餌劑有相同的二次時間趨勢,而百利普芬餌劑則只有一次時間效應顯著。就嘉義地區的試驗而言,處理前的蟻塚密度對於藥劑的防治率有顯著的影響,在時間效應上所有的藥劑處理皆有相同的時間趨勢。
Red imported fire ant (RIFA) arrived Taiwan around 2000 and has caused more or less medical, economic and ecological problems in the infested area. To evaluate the effectiveness of three bait-formulated insecticides, namely, Fipronil, Pyriproxyfen, and Spinosad, for managing RIFA, Taoyuan and Tainan District Agricultural Research and Extension Stations each conducted a filed experiment, respectively, in Taoyuan and Chiayi counties where infestation of RIFA was found. In both experiments, three 10x10 square meter plots were randomly assigned to each of the three insecticide treatments and a control treatment. Number of ant mounds on each plot was counted repeatedly during the first eight weeks after the application of treatments. From these longitudinal data, we can investigate four questions of interest: 1. Do these insecticides work in controlling RIFA? 2. Do these insecticides have the same effectiveness? 3. Does pre-treatment mound density influence the effects of the insecticides? 4. Does the evolution of ant mound density differ in time trend among treatments? This case study aims to compare the validity of different statistical procedures available in SAS/STAT (version 8) with respect to answering the above questions. Firstly, the relative mound density, i.e., ratio of mound number to the pre-treatment mound number, was log-transformed and plotted against time of measurement. It revealed that all these insecticides worked effectively in controlling RIFA and the every plot's profile of log-transformed relative mound density could be well approximated by a quadratic function of time. Secondly, Proc GLM was then invoked to perform a split plot in time analysis. It revealed that Pyriproxyfen was inferior to Fipronil and Spinosad in controlling RIFA and the time trend of declination of relative mound density differed significantly among insecticide treatments. The validity of this analysis was doubtful because the data of Taoyuan experiment failed to meet the Huynh-Feldt assumptions. Adding REPEATED statement to Proc GLM produced virtually the same results. Thirdly, the log-transformed relative mound density of each plot was separately fitted to a quadratic function of time, expressed in weeks after application of insecticides. Taking initial mound density as the covariate, covariance analyses were then used to investigate the effects of treatment groups on the plot-specific evolution over time. Only the Fipronil treatment in Taoyuan experiment showed significant effect in controlling RIFA. These results are invalid because the assumptions of covariance analysis were violated. Finally, we invoked Proc MIXED to fit a general linear mixed model to each of the two data sets. The results showed that all these insecticides were effective in both experiments. In Taoyuan experiment, the pre-treatment mound density gave no significant influence on the control rates of the insecticide. And quadratic time trend of the log-transformed relative mound density was detected in both the Fipronil and Spinosad treatments, while only linear time effect was significant in the Pyriproxyfen treatment. In Chiayi experiment, the pre-treatment mound density showed significant influence on the control rates of the insecticides. And all the three insecticides showed the same time trend in declining mound density.
URI: http://hdl.handle.net/11455/37128
Appears in Collections:農藝學系

文件中的檔案:

取得全文請前往華藝線上圖書館



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.