<p> The accuracy of a diagnostic test is often evaluated with the measures of sensitivity and specificity and the joint dependence between these two measures is captured by the receiver operating characteristic (ROC) curve. To combine multiple testing results from studies that are assumed to follow the same underlying probability law, a smooth summary receiver operating characteristic (SROC) curve can be fitted. Moses et al. (1993) proposed a least squares approach to fit the smooth SROC curve. </p> <p> In this thesis we overview the summary measures for the ROC curve in single study data as well as the summary statistics for the SROC curves in meta-analysis. These summary statistics include, the area under the curve (AUC), Q* statistic, area swept under the curve (ASC) and the partial area under the curve (pAUC). </p> <p> Our focus, however is mainly on the partial area under the SROC curve as it is being used frequently in meta-analysis of diagnostic testing. The appeal to use the pAUC instead of the full AUC is that the partial area can be used to focus on a clinically relevant region of the SROC curve where false positive rate (FPR) is small. Simulations and considerations for the use of the summary indices of the ROC and SROC curves are presented here. </p> / Thesis / Master of Science (MSc)

Receiver operating characteristic (ROC) curves are widely used in medical decision making. It was recognized in the last decade that only a specific region of the ROC curve is of clinical interest, which can be summarized by the partial area under the ROC curve (partial AUC). Early statistical methods for evaluating partial AUC assume that the data are from a specified underlying distribution. Nonparametric estimators of the partial AUC emerged recently, but there are theoretical issues to be addressed. In this dissertation, we propose two new nonparametric statistics, partially integrated ROC and partially integrated weighted ROC, for estimating partial AUC. We show that our partially integrated ROC statistic is a consistent estimator of the partial AUC, and derive its asymptotic distribution which is distribution free under the null hypothesis. In the partially integrated ROC statistic, when the ROC curve crosses the Uniform distribution function (CDF) and if the partial area evaluated contains the crossing point, or when there are multiple crossing, the partially integrated ROC statistic might not perform well. To address this issue, we propose the partially integrated weighted ROC statistic. This statistic evaluates the partially weighted AUC, where larger weight is given when the ROC curve is above the Uniform CDF and smaller weight is given when the ROC curve is below the Uniform CDF. We show that our partially integrated weighted ROC statistic is a consistent estimator of the partially weighted AUC. We derive its asymptotic distribution which is distribution free under the null hypothesis. We propose to apply our two nonparametric statistics to functional category analysis in microarray experiments. We define the functional category analysis to be the statistical identification of over-represented functional gene categories in a microarray experiment based on differential gene expression. We compare our statistics with existing methods for the functional category analysis both via simulation study and application to a real microarray data, and demonstrate that our two statistics are effective for identifying over-represented functional gene categories. We also emphasize the essential role of the empirical distribution function plots and the ROC curves in the functional category analysis.

本論文的研究目標是建構一個由多標記複合成的最佳疾病診斷工具，所考慮的評估準則為操作者特徵曲線在特定特異度範圍之線下面積(pAUC)。在常態分布假設下，我們推導多標記線性組合之pAUC以及最佳線性組合之必要條件。由於函數本身過於複雜使得計算困難。除此之外，我們也發現其最佳解可能不唯一，以及局部極值存在，這些情況使得現有演算法的運用受限，我們因此提出多重初始值演算法。當母體參數未知時，我們利用最大概似估計量以獲得樣本pAUC以及令其極大化之最佳線性組合，並證明樣本最佳線性組合將一致性地收斂到母體最佳線性組合。在進一步的研究中，我們針對單標記的邊際判別能力、多標記的複合判別能力以及個別標記的條件判別能力，分別提出相關統計檢定方法。這些統計檢定被運用至兩個標記選取的方法，分別是前進選擇法與後退淘汰法。我們運用這些方法以選取與疾病檢測有顯著相關的標記。本論文透過模擬研究來驗證所提出的演算法、統計檢定方法以及標記選取的方法。另外，也將這些方法運用在數組實際資料上。 / The aim of this work is to construct a composite diagnostic tool based on multiple biomarkers under the criterion of the partial area under a ROC curve (pAUC) for a predetermined specificity range. Recently several studies are interested in the optimal linear combination maximizing the whole area under a ROC curve (AUC). In this study, we focus on finding the optimal linear combination by a direct maximization of the pAUC under normal assumption. In order to find an analytic solution, the first derivative of the pAUC is derived. The form is so complicated, that a further validation on the Hessian matrix is difficult. In addition, we find that the pAUC maximizer may not be unique and sometimes, local maximizers exist. As a result, the existing algorithms, which depend on the initial-point, are inadequate to serve our needs. We propose a new algorithm by adopting several initial points at one time. In addition, when the population parameters are unknown and only a random sample data set is available, the maximizer of the sample version of the pAUC is shown to be a strong consistent estimator of its theoretical counterpart. We further focus on determining whether a biomarker set, or one specific biomarker has a significant contribution to the disease diagnosis. We propose three statistical tests for the identification of the discriminatory power. The proposed tests are applied to biomarker selection for reducing the variable number in advanced analysis. Numerical studies are performed to validate the proposed algorithm and the proposed statistical procedures.