American Statistician最新文献

When a researcher desires to test several treatment arms against a control arm, a two-stage adaptive design can be more efficient than a single-stage design where patients are equally allocated to all treatment arms and the control. We see this type of approach in clinical trials as a seamless Phase II - Phase III design. These designs require more statistical support and are less straightforward to plan and analyze than a standard single-stage design. To diminish the barriers associated with a Bayesian two-stage drop-the-losers design, we built a user-friendly point-and-click graphical user interface with R Shiny to aid researchers in planning such designs by allowing them to easily obtain trial operating characteristics, estimate statistical power and sample size, and optimize patient allocation in each stage to maximize power. We assume that endpoints are distributed normally with unknown but common variance between treatments. We recommend this software as an easy way to engage statisticians and researchers in two-stage designs as well as to actively investigate the power of two-stage designs relative to more traditional approaches. The software is freely available at https://github.com/stefangraw/Allocation-Power-Optimizer.

The Wilcoxon-Mann-Whitney (WMW) test is a popular rank-based two-sample testing procedure for the strong null hypothesis that the two samples come from the same distribution. A modified WMW test, the Fligner-Policello (FP) test, has been proposed for comparing the medians of two populations. A fact that may be underappreciated among some practitioners is that the FP test can also be used to test the strong null like the WMW. In this paper we compare the power of the WMW and FP tests for testing the strong null. Our results show that neither test is uniformly better than the other and that there can be substantial differences in power between the two choices. We propose a new, modified WMW test that combines the WMW and FP tests. Monte Carlo studies show that the combined test has good power compared to either the WMW and FP test. We provide a fast implementation of the proposed test in an open-source software. Supplementary materials are available online.

The score statistic continues to be a fundamental tool for statistical inference. In the analysis of data from high-throughput genomic assays, inference on the basis of the score usually enjoys greater stability, considerably higher computational efficiency, and lends itself more readily to the use of resampling methods than the asymptotically equivalent Wald or likelihood ratio tests. The score function often depends on a set of unknown nuisance parameters which have to be replaced by estimators, but can be improved by calculating the efficient score, which accounts for the variability induced by estimating these parameters. Manual derivation of the efficient score is tedious and error-prone, so we illustrate using computer algebra to facilitate this derivation. We demonstrate this process within the context of a standard example from genetic association analyses, though the techniques shown here could be applied to any derivation, and have a place in the toolbox of any modern statistician. We further show how the resulting symbolic expressions can be readily ported to compiled languages, to develop fast numerical algorithms for high-throughput genomic analysis. We conclude by considering extensions of this approach. The code featured in this report is available online as part of the supplementary material.

Demand for data science education is surging and traditional courses offered by statistics departments are not meeting the needs of those seeking training. This has led to a number of opinion pieces advocating for an update to the Statistics curriculum. The unifying recommendation is that computing should play a more prominent role. We strongly agree with this recommendation, but advocate the main priority is to bring applications to the forefront as proposed by Nolan and Speed (1999). We also argue that the individuals tasked with developing data science courses should not only have statistical training, but also have experience analyzing data with the main objective of solving real-world problems. Here, we share a set of general principles and offer a detailed guide derived from our successful experience developing and teaching a graduate-level, introductory data science course centered entirely on case studies. We argue for the importance of statistical thinking, as defined by Wild and Pfannkuch (1999) and describe how our approach teaches students three key skills needed to succeed in data science, which we refer to as creating, connecting, and computing. This guide can also be used for statisticians wanting to gain more practical knowledge about data science before embarking on teaching an introductory course.

Within the statistics community, a number of guiding principles for sharing data have emerged; however, these principles are not always made clear to collaborators generating the data. To bridge this divide, we have established a set of guidelines for sharing data. In these, we highlight the need to provide raw data to the statistician, the importance of consistent formatting, and the necessity of including all essential experimental information and pre-processing steps carried out to the statistician. With these guidelines we hope to avoid errors and delays in data analysis.

We consider settings where it is of interest to fit and assess regression submodels that arise as various explanatory variables are excluded from a larger regression model. The larger model is referred to as the full model; the submodels are the reduced models. We show that a computationally efficient approximation to the regression estimates under any reduced model can be obtained from a simple weighted least squares (WLS) approach based on the estimated regression parameters and covariance matrix from the full model. This WLS approach can be considered an extension to unbiased estimating equations of a first-order Taylor series approach proposed by Lawless and Singhal. Using data from the 2010 Nationwide Inpatient Sample (NIS), a 20% weighted, stratified, cluster sample of approximately 8 million hospital stays from approximately 1000 hospitals, we illustrate the WLS approach when fitting interval censored regression models to estimate the effect of type of surgery (robotic versus nonrobotic surgery) on hospital length-of-stay while adjusting for three sets of covariates: patient-level characteristics, hospital characteristics, and zip-code level characteristics. Ordinarily, standard fitting of the reduced models to the NIS data takes approximately 10 hours; using the proposed WLS approach, the reduced models take seconds to fit.

Collaborative biostatistics faculties (CBF) are increasingly valued by academic health centers (AHCs) for their role in increasingsuccess rates of grants and publications, and educating medical students and clinical researchers. Some AHCs have a biostatistics department that consists of only biostatisticians focused on methodological research, collaborative research, and education. Others may have a biostatistics unit within an interdisciplinary department, or statisticians recruited into clinical departments. Within each model, there is also variability in environment, influenced by the chair's background, research focus of colleagues, type of students taught, funding sources, and whether the department is in a medical school or school of public health. CBF appointments may be tenure track or non-tenure, and expectations for promotion may vary greatly depending on the type of department, track, and the AHC. In this article, the authors identify strategies for developing early-stage CBFs in four domains: 1)Influenceof department/environment, 2) Skills to develop, 3) Ways to increase productivity, and 4) Ways to document accomplishments. Graduating students and postdoctoral fellows should consider the first domain when choosing a faculty position. Early-stage CBFs will benefit by understanding the requirements of their environment early in their appointment and by modifying the provided progression grid with their chair and mentoring team as needed. Following this personalized grid will increase the chances of a satisfying career with appropriate recognition for academic accomplishments.

Although propensity scores have been central to the estimation of causal effects for over 30 years, only recently has the statistical literature begun to consider in detail methods for Bayesian estimation of propensity scores and causal effects. Underlying this recent body of literature on Bayesian propensity score estimation is an implicit discordance between the goal of the propensity score and the use of Bayes theorem. The propensity score condenses multivariate covariate information into a scalar to allow estimation of causal effects without specifying a model for how each covariate relates to the outcome. Avoiding specification of a detailed model for the outcome response surface is valuable for robust estimation of causal effects, but this strategy is at odds with the use of Bayes theorem, which presupposes a full probability model for the observed data that adheres to the likelihood principle. The goal of this paper is to explicate this fundamental feature of Bayesian estimation of causal effects with propensity scores in order to provide context for the existing literature and for future work on this important topic.

There is growing frustration with the concept of the p-value. Besides having an ambiguous interpretation, the p-value can be made as small as desired by increasing the sample size, n. The p-value is outdated and does not make sense with big data: Everything becomes statistically significant. The root of the problem with the p-value is in the mean comparison. We argue that statistical uncertainty should be measured on the individual, not the group, level. Consequently, standard deviation (SD), not standard error (SE), error bars should be used to graphically present the data on two groups. We introduce a new measure based on the discrimination of individuals/objects from two groups, and call it the D-value. The D-value can be viewed as the n-of-1 p-value because it is computed in the same way as p while letting n equal 1. We show how the D-value is related to discrimination probability and the area above the receiver operating characteristic (ROC) curve. The D-value has a clear interpretation as the proportion of patients who get worse after the treatment, and as such facilitates to weigh up the likelihood of events under different scenarios. [Received January 2015. Revised June 2015.].