Operating characteristics and extensions of the false discovery rate procedure

Summary. We investigate the operating characteristics of the Benjamini–Hochberg false discovery rate procedure for multiple testing. This is a distribution-free method that controls the expected fraction of falsely rejected null hypotheses among those rejected. The paper provides a framework for understanding more about this procedure. We first study the asymptotic properties of the `deciding point' D that determines the critical p -value. From this, we obtain explicit asymptotic expressions for a particular risk function. We introduce the dual notion of false non-rejections and we consider a risk function that combines the false discovery rate and false non-rejections. We also consider the optimal procedure with respect to a measure of conditional risk.     

A note on the adaptive control of false discovery rates

Summary.  The use of a fixed rejection region for multiple hypothesis testing has been shown to outperform standard fixed error rate approaches when applied to control of the false discovery rate. In this work it is demonstrated that, if the original step-up procedure of Benjamini and Hochberg is modified to exercise adaptive control of the false discovery rate, its performance is virtually identical to that of the fixed rejection region approach. In addition, the dependence of both methods on the proportion of true null hypotheses is explored, with a focus on the difficulties that are involved in the estimation of this quantity.     

Estimating false discovery proportion in multiple comparison under dependency

Multiple comparison, which tests tens of thousands hypotheses simultaneously, has been developed extensively under the assumption of independence of test statistics. Practically the independence assumption, without which multiple comparison inference becomes very challenging, may not be suitable. Fan, Han and Gu proposed an efficient way to estimate the false discovery proportion when the test statistics are normally distributed with arbitrary covariance dependence. The normal distribution assumption constrains the employment of their estimator. In this article, we generalize their method to the cases where the test statistics are built on the normally distributed sample.     

A direct approach to false discovery rates

Summary. Multiple-hypothesis testing involves guarding against much more complicated errors than single-hypothesis testing. Whereas we typically control the type I error rate for a single-hypothesis test, a compound error rate is controlled for multiple-hypothesis tests. For example, controlling the false discovery rate FDR traditionally involves intricate sequential p -value rejection methods based on the observed data. Whereas a sequential p -value method fixes the error rate and estimates its corresponding rejection region, we propose the opposite approach—we fix the rejection region and then estimate its corresponding error rate. This new approach offers increased applicability, accuracy and power. We apply the methodology to both the positive false discovery rate pFDR and FDR, and provide evidence for its benefits. It is shown that pFDR is probably the quantity of interest over FDR. Also discussed is the calculation of the q -value, the pFDR analogue of the p -value, which eliminates the need to set the error rate beforehand as is traditionally done. Some simple numerical examples are presented that show that this new approach can yield an increase of over eight times in power compared with the Benjamini–Hochberg FDR method.     

A modified false discovery rate multiple-comparisons procedure for discrete data, applied to human immunodeficiency virus genetics

Summary.  To help to design vaccines for acquired immune deficiency syndrome that protect broadly against many genetic variants of the human immunodeficiency virus, the mutation rates at 118 positions in HIV amino-acid sequences of subtype C versus those of subtype B were compared. The false discovery rate (FDR) multiple-comparisons procedure can be used to determine statistical significance. When the test statistics have discrete distributions, the FDR procedure can be made more powerful by a simple modification. The paper develops a modified FDR procedure for discrete data and applies it to the human immunodeficiency virus data. The new procedure detects 15 positions with significantly different mutation rates compared with 11 that are detected by the original FDR method. Simulations delineate conditions under which the modified FDR procedure confers large gains in power over the original technique. In general FDR adjustment methods can be improved for discrete data by incorporating the modification proposed.     

An exponentially weighted moving average chart controlling false discovery rate

A new control chart is developed by using the exponentially weighted moving average (EWMA) statistics and a multiple testing procedure for controlling false discovery rate. The multiple testing procedure considers not only the current EWMA statistic, but also a given number of previous statistics at the same time. Numerical simulations are accomplished to evaluate the performance of the proposed control chart in terms of the average run length and the conditional expected delay. The results are compared with those of the existing control charts including the X-bar chart, EWMA, and cumulative sum control charts. Case studies with real data-sets are also presented.     

False discovery rates for large-scale model checking under certain dependence

In many scientific fields, it is interesting and important to determine whether an observed data stream comes from a prespecified model or not, particularly when the number of data streams is of large scale, where multiple hypotheses testing is necessary. In this article, we consider large-scale model checking under certain dependence among different data streams observed at the same time. We propose a false discovery rate (FDR) control procedure to check those unusual data streams. Specifically, we derive an approximation of false discovery and construct a point estimate of FDR. Theoretical results show that, under some mild assumptions, our proposed estimate of FDR is simultaneously conservatively consistent with the true FDR, and hence it is an asymptotically strong control procedure. Simulation comparisons with some competing procedures show that our proposed FDR procedure behaves better in general settings. Application of our proposed FDR procedure is illustrated by the StarPlus fMRI data.     

Power and stability comparisons of multiple testing procedures with false discovery rate control

High-throughput data analyses are widely used for examining differential gene expression, identifying single nucleotide polymorphisms, and detecting methylation loci. False discovery rate (FDR) has been considered a proper type I error rate to control for discovery-based high-throughput data analysis. Various multiple testing procedures have been proposed to control the FDR. The power and stability properties of some commonly used multiple testing procedures have not been extensively investigated yet, however. Simulation studies were conducted to compare power and stability properties of five widely used multiple testing procedures at different proportions of true discoveries for various sample sizes for both independent and dependent test statistics. Storey's two linear step-up procedures showed the best performance among all tested procedures considering FDR control, power, and variance of true discoveries. Leukaemia and ovarian cancer microarray studies were used to illustrate the power and stability characteristics of these five multiple testing procedures with FDR control.     

A Bayesian discovery procedure

Summary.  We discuss a Bayesian discovery procedure for multiple-comparison problems. We show that, under a coherent decision theoretic framework, a loss function combining true positive and false positive counts leads to a decision rule that is based on a threshold of the posterior probability of the alternative. Under a semiparametric model for the data, we show that the Bayes rule can be approximated by the optimal discovery procedure, which was recently introduced by Storey. Improving the approximation leads us to a Bayesian discovery procedure, which exploits the multiple shrinkage in clusters that are implied by the assumed non-parametric model. We compare the Bayesian discovery procedure and the optimal discovery procedure estimates in a simple simulation study and in an assessment of differential gene expression based on microarray data from tumour samples. We extend the setting of the optimal discovery procedure by discussing modifications of the loss function that lead to different single-thresholding statistics. Finally, we provide an application of the previous arguments to dependent (spatial) data.     

More Powerful Control of the False Discovery Rate Under Dependence

In a breakthrough paper, Benjamini and Hochberg (J Roy Stat Soc Ser B 57:289–300, 1995) proposed a new error measure for multiple testing, the FDR; and developed a distribution-free procedure to control it under independence among the test statistics. In this paper we argue by extensive simulation and theoretical considerations that the assumption of independence is not needed. Along the lines of (Ann Stat 32:1035–1061, 2004b), we moreover provide a more powerful method, that exploits an estimator of the number of false nulls among the tests. We propose a whole family of iterative estimators that prove robust under dependence and independence between the test statistics. These estimators can be used to improve also classical multiple testing procedures, and in general to estimate the weight of a known component in a mixture distribution. Innovations are illustrated by simulations.     

Variance of the number of false discoveries

Summary.  In high throughput genomic work, a very large number d of hypotheses are tested based on n ≪ d data samples. The large number of tests necessitates an adjustment for false discoveries in which a true null hypothesis was rejected. The expected number of false discoveries is easy to obtain. Dependences between the hypothesis tests greatly affect the variance of the number of false discoveries. Assuming that the tests are independent gives an inadequate variance formula. The paper presents a variance formula that takes account of the correlations between test statistics. That formula involves O ( d ²) correlations, and so a naïve implementation has cost O ( nd ²). A method based on sampling pairs of tests allows the variance to be approximated at a cost that is independent of d .     

False discovery rate control for non-positively regression dependent test statistics

In this paper we present a modification of the Benjamini and Hochberg false discovery rate controlling procedure for testing non-positive dependent test statistics. The new testing procedure makes use of the same series of linearly increasing critical values. Yet, in the new procedure the set of p-values is divided into subsets of positively dependent p-values, and each subset of p-values is separately sorted and compared to the series of critical values. In the first part of the paper we introduce the new testing methodology, discuss the technical issues needed to apply the new approach, and apply it to data from a genetic experiment.     

Evaluation of false discovery rate and power via sample size in microarray studies

Microarray studies are now common for human, agricultural plant and animal studies. False discovery rate (FDR) is widely used in the analysis of large-scale microarray data to account for problems associated with multiple testing. A well-designed microarray study should have adequate statistical power to detect the differentially expressed (DE) genes, while keeping the FDR acceptably low. In this paper, we used a mixture model of expression responses involving DE genes and non-DE genes to analyse theoretical FDR and power for simple scenarios where it is assumed that each gene has equal error variance and the gene effects are independent. A simulation study was used to evaluate the empirical FDR and power for more complex scenarios with unequal error variance and gene dependence. Based on this approach, we present a general guide for sample size requirement at the experimental design stage for prospective microarray studies. This paper presented an approach to explicitly connect the sample size with FDR and power. While the methods have been developed in the context of one-sample microarray studies, they are readily applicable to two-sample, and could be adapted to multiple-sample studies.     

Controlling type I error rates in multi-arm clinical trials: A case for the false discovery rate

Multi-arm trials are an efficient way of simultaneously testing several experimental treatments against a shared control group. As well as reducing the sample size required compared to running each trial separately, they have important administrative and logistical advantages. There has been debate over whether multi-arm trials should correct for the fact that multiple null hypotheses are tested within the same experiment. Previous opinions have ranged from no correction is required, to a stringent correction (controlling the probability of making at least one type I error) being needed, with regulators arguing the latter for confirmatory settings. In this article, we propose that controlling the false-discovery rate (FDR) is a suitable compromise, with an appealing interpretation in multi-arm clinical trials. We investigate the properties of the different correction methods in terms of the positive and negative predictive value (respectively how confident we are that a recommended treatment is effective and that a non-recommended treatment is ineffective). The number of arms and proportion of treatments that are truly effective is varied. Controlling the FDR provides good properties. It retains the high positive predictive value of FWER correction in situations where a low proportion of treatments is effective. It also has a good negative predictive value in situations where a high proportion of treatments is effective. In a multi-arm trial testing distinct treatment arms, we recommend that sponsors and trialists consider use of the FDR.     

A note on estimating the false discovery rate under mixture model

In this note, we focus on estimating the false discovery rate (FDR) of a multiple testing method with a common, non-random rejection threshold under a mixture model. We develop a new class of estimates of the FDR and prove that it is less conservatively biased than what is traditionally used. Numerical evidence is presented to show that the mean squared error (MSE) is also often smaller for the present class of estimates, especially in small-scale multiple testings. A similar class of estimates of the positive false discovery rate (pFDR) less conservatively biased than what is usually used is then proposed. When modified using our estimate of the pFDR and applied to a gene-expression data, Storey's q-value method identifies a few more significant genes than his original q-value method at certain thresholds. The BH like method developed by thresholding our estimate of the FDR is shown to control the FDR in situations where the p  -values have the same dependence structure as required by the BH method and, for lack of information about the proportion _π0${\pi }_0$ of true null hypotheses, it is reasonable to assume that _π0${\pi }_0$ is uniformly distributed over (0,1).     

A simple procedure for the selection of significant effects

Summary.  Given a large number of test statistics, a small proportion of which represent departures from the relevant null hypothesis, a simple rule is given for choosing those statistics that are indicative of departure. It is based on fitting by moments a mixture model to the set of test statistics and then deriving an estimated likelihood ratio. Simulation suggests that the procedure has good properties when the departure from an overall null hypothesis is not too small.     

Bayesian False Discovery Rate Wavelet Shrinkage: Theory and Applications

Statistical inference in the wavelet domain remains a vibrant area of contemporary statistical research because of desirable properties of wavelet representations and the need of scientific community to process, explore, and summarize massive data sets. Prime examples are biomedical, geophysical, and internet related data. We propose two new approaches to wavelet shrinkage/thresholding.

In the spirit of Efron and Tibshirani's recent work on local false discovery rate, we propose Bayesian Local False Discovery Rate (BLFDR), where the underlying model on wavelet coefficients does not assume known variances. This approach to wavelet shrinkage is shown to be connected with shrinkage based on Bayes factors. The second proposal, Bayesian False Discovery Rate (BaFDR), is based on ordering of posterior probabilities of hypotheses on true wavelets coefficients being null, in Bayesian testing of multiple hypotheses.

We demonstrate that both approaches result in competitive shrinkage methods by contrasting them to some popular shrinkage techniques.