Evaluation of the statistical power for multiple tests: a case study

It is challenging to estimate the statistical power when a complicated testing strategy is used to adjust for the type-I error for multiple comparisons in a clinical trial. In this paper, we use the Bonferroni Inequality to estimate the lower bound of the statistical power assuming that test statistics are approximately normally distributed and the correlation structure among test statistics is unknown or only partially known. The method was applied to the design of a clinical study for sample size and statistical power estimation.     

A power study of goodness-of-fit tests for multivariate normality implemented in R

Multivariate statistical analysis procedures often require data to be multivariate normally distributed. Many tests have been developed to verify if a sample could indeed have come from a normally distributed population. These tests do not all share the same sensitivity for detecting departures from normality, and thus a choice of test is of central importance. This study investigates through simulated data the power of those tests for multivariate normality implemented in the statistic software R and pits them against the variant of testing each marginal distribution for normality. The results of testing two-dimensional data at a level of significance α=5% showed that almost one-third of those tests implemented in R do not have a type I error below this. Other tests outperformed the naive variant in terms of power even when the marginals were not normally distributed. Even though no test was consistently better than all alternatives with every alternative distribution, the energy-statistic test always showed relatively good power across all tested sample sizes.     

Power and sample size when multiple endpoints are considered

A common approach to analysing clinical trials with multiple outcomes is to control the probability for the trial as a whole of making at least one incorrect positive finding under any configuration of true and false null hypotheses. Popular approaches are to use Bonferroni corrections or structured approaches such as, for example, closed-test procedures. As is well known, such strategies, which control the family-wise error rate, typically reduce the type I error for some or all the tests of the various null hypotheses to below the nominal level. In consequence, there is generally a loss of power for individual tests. What is less well appreciated, perhaps, is that depending on approach and circumstances, the test-wise loss of power does not necessarily lead to a family wise loss of power. In fact, it may be possible to increase the overall power of a trial by carrying out tests on multiple outcomes without increasing the probability of making at least one type I error when all null hypotheses are true. We examine two types of problems to illustrate this. Unstructured testing problems arise typically (but not exclusively) when many outcomes are being measured. We consider the case of more than two hypotheses when a Bonferroni approach is being applied while for illustration we assume compound symmetry to hold for the correlation of all variables. Using the device of a latent variable it is easy to show that power is not reduced as the number of variables tested increases, provided that the common correlation coefficient is not too high (say less than 0.75). Afterwards, we will consider structured testing problems. Here, multiplicity problems arising from the comparison of more than two treatments, as opposed to more than one measurement, are typical. We conduct a numerical study and conclude again that power is not reduced as the number of tested variables increases.     

P-value adjustment for multiple binary endpoints

The p-value-based adjustment of individual endpoints and the global test for an overall inference are the two general approaches for the analysis of multiple endpoints. Statistical procedures developed for testing multivariate outcomes often assume that the multivariate endpoints are either independent or normally distributed. This paper presents a general approach for the analysis of multivariate binary data under the framework of generalized linear models. The generalized estimating equations (GEE) approach is applied to estimate the correlation matrix of the test statistics using the identity and exchangeable working correlation matrices with the model-based as well as robust estimators. The objectives of the approaches are the adjustment of p-values of individual endpoints to identify the affected endpoints as well as the global test of an overall effect. A Monte Carlo simulation was conducted to evaluate the overall family wise error (FWE) rates of the single-step down p-value adjustment approach from two adjustment methods to three global test statistics. The p-value adjustment approach seems to control the FWE better than the global approach Applications of the proposed methods are illustrated by analyzing a carcinogenicity experiment designed to study the dose response trend for 10 tumor sites, and a developmental toxicity experiment with three malformation types: external, visceral, and skeletal.     

Resampling-based stepwise multiple testing procedures with applications to clinical trial data

Endpoints in clinical trials are often highly correlated. However, the commonly used multiple testing procedures in clinical trials either do not take into consideration the correlations among test statistics or can only exploit known correlations. Westfall and Young constructed a resampling-based stepdown method that implicitly utilizes the correlation structure of test statistics in situations with unknown correlations. However, their method requires a “subset pivotality” assumption. Romano and Wolf proposed a more general stepdown method, which does not require such an assumption. There is at present little experience with the application of such methods in analyzing clinical trial data. We advocate the application of resampling-based multiple testing procedures to clinical trials data when appropriate. We have conjectured that the resampling-based stepdown methods can be extended to a stepup procedure under appropriate assumptions and examined the performance of both stepdown and stepup methods under a variety of correlation structures and distribution types. Results from our simulation studies support the use of the resampling-based methods under various scenarios, including binary data and small samples, with strong control of Family wise type I error rate (FWER). Under positive dependence and for binary data even under independence, the resampling-based methods are more powerful than the Holm and Hochberg methods. Last, we illustrate the advantage of the resampling-based stepwise methods with two clinical trial data examples: a cardiovascular outcome trial and an oncology trial.     

A flexible fixed-sequence testing method for hierarchically ordered correlated multiple endpoints in clinical trials

Statistical approaches for addressing multiplicity in clinical trials range from the very conservative (the Bonferroni method) to the least conservative the fixed sequence approach. Recently, several authors proposed methods that combine merits of the two extreme approaches. Wiens [2003. A fixed sequence Bonferroni procedure for testing multiple endpoints. Pharmaceutical Statist. 2003, 2, 211–215], for example, considered an extension of the Bonferroni approach where the type I error rate (α)$(\alpha )$ is allocated among the endpoints, however, testing proceeds in a pre-determined order allowing the type I error rate to be saved for later use as long as the null hypotheses are rejected. This leads to a higher power of the test in testing later null hypotheses. In this paper, we consider an extension of Wiens’ approach by taking into account correlations among endpoints for achieving higher flexibility in testing. We show strong control of the family-wise type I error rate for this extension and provide critical values and significance levels for testing up to three endpoints with equal correlations and show how to calculate them for other correlation structures. We also present results of a simulation experiment for comparing the power of the proposed method with those of Wiens’ and others. The results of this experiment show that the magnitude of the gain in power of the proposed method depends on the prospective ordering of testing of the endpoints, the magnitude of the treatment effects of the endpoints and the magnitude of correlation between endpoints. Finally, we consider applications of the proposed method for clinical trials with multiple time points and multiple doses, where correlations among endpoints frequently arise.     

A unified sequential test procedure for simultaneous testing the equality of several binomial proportions to a specified standard

In this article, we propose a unified sequentially rejective test procedure for testing simultaneously the equality of several independent binomial proportions to a specified standard. The proposed test procedure is general enough to include some well-known multiple testing procedures such as the Ordinary Bonferroni procedure, Hochberg procedure and Rom procedure. It involves multiple tests of significance based on the simple binomial tests (exact or approximate) which can be easily found in many elementary standard statistics textbooks. Unlike the traditional Chi-square test of the overall hypothesis, the procedure can identify the subset of the binomial proportions, which are different from the prespecified standard with the control of the familywise type I error rate. Moreover, the power computation of the procedure is provided and the procedure is illustrated by two real examples from an ecological study and a carcinogenicity study.     

Inflated statistical significance of student's t test associated with small intersubject correlation

The independence assumption in statistical significance testing becomes increasingly crucial and unforgiving as sample size increases. Seemingly, inconsequential violations of this assumption can substantially increase the probability of a Type I error if sample sizes are large. In the case of Student's t test, it is found that correlations within samples in a range from 0.01 to 0.05 can lead to rejection of a true null hypothesis with high probability, if N is 50, 100 or larger.     

A Necessary Power Divergence Type Family Tests of Multivariate Normality

In a recent article, Cardoso de Oliveira and Ferreira have proposed a multivariate extension of the univariate chi-squared normality test, using a known result for the distribution of quadratic forms in normal variables. In this article, we propose a family of power divergence type test statistics for testing the hypothesis of multinormality. The proposed family of test statistics includes as a particular case the test proposed by Cardoso de Oliveira and Ferreira. We assess the performance of the new family of test statistics by using Monte Carlo simulation. In this context, the type I error rates and the power of the tests are studied, for important family members. Moreover, the performance of significant members of the proposed test statistics are compared with the respective performance of a multivariate normality test, proposed recently by Batsidis and Zografos. Finally, two well-known data sets are used to illustrate the method developed in this article as well as the specialized test of multivariate normality proposed by Batsidis and Zografos.     

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.     

Bayesian adaptive bandit-based designs using the Gittins index for multi-armed trials with normally distributed endpoints

Adaptive designs for multi-armed clinical trials have become increasingly popular recently because of their potential to shorten development times and to increase patient response. However, developing response-adaptive designs that offer patient-benefit while ensuring the resulting trial provides a statistically rigorous and unbiased comparison of the different treatments included is highly challenging. In this paper, the theory of Multi-Armed Bandit Problems is used to define near optimal adaptive designs in the context of a clinical trial with a normally distributed endpoint with known variance. We report the operating characteristics (type I error, power, bias) and patient-benefit of these approaches and alternative designs using simulation studies based on an ongoing trial. These results are then compared to those recently published in the context of Bernoulli endpoints. Many limitations and advantages are similar in both cases but there are also important differences, specially with respect to type I error control. This paper proposes a simulation-based testing procedure to correct for the observed type I error inflation that bandit-based and adaptive rules can induce.     

An Adaptive Method of Variable Selection in Regression

An adaptive variable selection procedure is proposed which uses an adaptive test along with a stepwise procedure to select variables for a multiple regression model. We compared this adaptive stepwise procedure to methods that use Akaike's information criterion, Schwartz's information criterion, and Sawa's information criterion. The simulation studies demonstrated that the adaptive stepwise method is more effective than the traditional variable selection methods if the error distribution is not normally distributed. If the error distribution is known to be normally distributed, the variable selection method based on Sawa's information criteria appears to be superior to the other methods. Unless the error distribution is known to be normally distributed, the adaptive stepwise method is recommended.     

Nonparametric multiple comparison procedures for unbalanced one-way factorial designs

In this paper, we present several nonparametric multiple comparison (MC) procedures for unbalanced one-way factorial designs. The nonparametric hypotheses are formulated by using normalized distribution functions and the comparisons are carried out on the basis of the relative treatment effects. The proposed test statistics take the form of linear pseudo rank statistics and the asymptotic joint distribution of the pseudo rank statistics for testing treatments versus control satisfies the multivariate totally positive of order two condition irrespective of the correlations among the rank statistics. Therefore, in the context of MCs of treatments versus control, the nonparametric Simes test is validated for the global testing of the intersection hypothesis. For simultaneous testing of individual hypotheses, the nonparametric Hochberg stepup procedure strongly controls the familywise type I error rate asymptotically. With regard to all pairwise comparisons, we generalize various single-step and stagewise procedures to perform comparisons on the relative treatment effects. To further compare with normal theory counterparts, the asymptotic relative efficiencies of the nonparametric MC procedures with respect to the parametric MC procedures are derived under a sequence of Pitman alternatives in a nonparametric location shift model for unbalanced one-way layouts. Monte Carlo simulations are conducted to demonstrate the validity and power of the proposed nonparametric MC procedures.     

Testing hypotheses on coefficients of variation from a series of two-armed experiments

We consider the problem of testing hypotheses on the difference of the coefficients of variation from several two-armed experiments with normally distributed outcomes. In particular, we deal with testing the homogeneity of the difference of the coefficients of variation and testing the equality of the difference of the coefficients of variation to a specified value. The test statistics proposed are derived in a limiting one-way classification with fixed effects and heteroscedastic error variances, using results from analysis of variance. By way of simulation, the performance of these test statistics is compared for both testing problems considered.     

STEPWISE TESTS WHEN THE TEST STATISTICS ARE INDEPENDENT

Consider the problem of simultaneously testing a nonhierarchical finite family of hypotheses based on independent test statistics. A general stepwise test is defined, of which the well known step-down and step-up tests are special cases. The step-up test is shown to dominate the other stepwise tests, including the step-down test, for situations of practical importance. When testing against two-sided alternatives, it is pointed out that if the step-up test is augmented to include directional decisions then the augmented step-up test controls the type I and III familywise error jointly at the original level q. The definition of the adjusted p values for the step-up test is justified. The results are illustrated by a numerical example.     

Robust inference from multiple test statistics via permutations: a better alternative to the single test statistic approach for randomized trials

Formal inference in randomized clinical trials is based on controlling the type I error rate associated with a single pre‐specified statistic. The deficiency of using just one method of analysis is that it depends on assumptions that may not be met. For robust inference, we propose pre‐specifying multiple test statistics and relying on the minimum p‐value for testing the null hypothesis of no treatment effect. The null hypothesis associated with the various test statistics is that the treatment groups are indistinguishable. The critical value for hypothesis testing comes from permutation distributions. Rejection of the null hypothesis when the smallest p‐value is less than the critical value controls the type I error rate at its designated value. Even if one of the candidate test statistics has low power, the adverse effect on the power of the minimum p‐value statistic is not much. Its use is illustrated with examples. We conclude that it is better to rely on the minimum p‐value rather than a single statistic particularly when that single statistic is the logrank test, because of the cost and complexity of many survival trials. Copyright © 2013 John Wiley & Sons, Ltd.     

Robust 2k factorial design with Weibull error distributions

It is well known that the least squares method is optimal only if the error distributions are normally distributed. However, in practice, non-normal distributions are more prevalent. If the error terms have a non-normal distribution, then the efficiency of least squares estimates and tests is very low. In this paper, we consider the 2^k factorial design when the distribution of error terms are Weibull W(p,σ). From the methodology of modified likelihood, we develop robust and efficient estimators for the parameters in 2^k factorial design. F statistics based on modified maximum likelihood estimators (MMLE) for testing the main effects and interaction are defined. They are shown to have high powers and better robustness properties as compared to the normal theory solutions. A real data set is analysed.     

Tests of additivity in mixed and fixed effect two-way ANOVA models with single sub-class numbers

In variety testing as well as in psychological assessment, the situation occurs that in a two-way ANOVA-type model with only one replication per cell, analysis is done under the assumption of no interaction between the two factors. Tests for this situation are known only for fixed factors and normally distributed outcomes. In the following we will present five additivity tests and apply them to fixed and mixed models and to quantitative as well as to Bernoulli distributed data. We consider their performance via simulation studies with respect to the type-I-risk and power. Furthermore, two new approaches will be presented, one being a modification of Tukey’s test and the other being a new experimental design to test for interactions.     

Testing variance components in balanced linear growth curve models

It is well known that the testing of zero variance components is a non-standard problem since the null hypothesis is on the boundary of the parameter space. The usual asymptotic chi-square distribution of the likelihood ratio and score statistics under the null does not necessarily hold because of this null hypothesis. To circumvent this difficulty in balanced linear growth curve models, we introduce an appropriate test statistic and suggest a permutation procedure to approximate its finite-sample distribution. The proposed test alleviates the necessity of any distributional assumptions for the random effects and errors and can easily be applied for testing multiple variance components. Our simulation studies show that the proposed test has Type I error rate close to the nominal level. The power of the proposed test is also compared with the likelihood ratio test in the simulations. An application on data from an orthodontic study is presented and discussed.