Sensitivity of the fractional Bayes factor to prior distributions

The authors derive a measure of the sensitivity of the fractional Bayes factor, an index which is used to compare models when the priors for their respective parameters are improper, or when there is concern about robustness of the prior specification. They prove that in a large class of problems, this measure is a decreasing function of the fraction of the sample used to update the prior distribution before the models are compared.     

A note on the difference between two likelihood based methods in growth curves

A new result is proved on the difference between ML and REML in the classical growth curve model, concerning the estimated variances of the regression coefficients. Simulations indicate that REML provides estimated variances closer to their true corresponding values, giving confidence intervals which are not misleadingly short.     

Robustness and Corrections for Sample Size Adaptation Strategies Based on Effect Size Estimation

A study on the robustness of the adaptation of the sample size for a phase III trial on the basis of existing phase II data is presented—when phase III is lower than phase II effect size. A criterion of clinical relevance for phase II results is applied in order to launch phase III, where data from phase II cannot be included in statistical analysis. The adaptation consists in adopting the conservative approach to sample size estimation, which takes into account the variability of phase II data. Some conservative sample size estimation strategies, Bayesian and frequentist, are compared with the calibrated optimal γ conservative strategy (viz. COS) which is the best performer when phase II and phase III effect sizes are equal. The Overall Power (OP) of these strategies and the mean square error (MSE) of their sample size estimators are computed under different scenarios, in the presence of the structural bias due to lower phase III effect size, for evaluating the robustness of the strategies. When the structural bias is quite small (i.e., the ratio of phase III to phase II effect size is greater than 0.8), and when some operating conditions for applying sample size estimation hold, COS can still provide acceptable results for planning phase III trials, even if in bias absence the OP was higher.

Main results concern the introduction of a correction, which affects just sample size estimates and not launch probabilities, for balancing the structural bias. In particular, the correction is based on a postulation of the structural bias; hence, it is more intuitive and easier to use than those based on the modification of Type I or/and Type II errors. A comparison of corrected conservative sample size estimation strategies is performed in the presence of a quite small bias. When the postulated correction is right, COS provides good OP and the lowest MSE. Moreover, the OPs of COS are even higher than those observed without bias, thanks to higher launch probability and a similar estimation performance. The structural bias can therefore be exploited for improving sample size estimation performances. When the postulated correction is smaller than necessary, COS is still the best performer, and it also works well. A higher than necessary correction should be avoided.     

Influential Observations in Time Series

This article studies how to identify influential observations in univariate autoregressive integrated moving average time series models and how to measure their effects on the estimated parameters of the model. The sensitivity of the parameters to the presence of either additive or innovational outliers is analyzed, and influence statistics based on the Mahalanobis distance are presented. The statistic linked to additive outliers is shown to be very useful for indicating the robustness of the fitted model to the given data set. Its application is illustrated using a relevant set of historical data.     

Robust discriminant analysis: Training data breakdown point

This paper discusses the robustness of discriminant analysis against contamination in the training data, the test data are assumed uncontaminated. The concept of training data breakdown point for discriminant analysis is introduced. It is quite different from the usual breakdown point in robust statistics. In the robust location parameter estimation problem, outliers are the main concern, but in discriminant analysis, not only are outliers a concern, but also inliers.     

Non-crossing large-margin probability estimation and its application to robust SVM via preconditioning

Many large-margin classifiers such as the Support Vector Machine (SVM) sidestep estimating conditional class probabilities and target the discovery of classification boundaries directly. However, estimation of conditional class probabilities can be useful in many applications. Wang, Shen, and Liu (2008) bridged the gap by providing an interval estimator of the conditional class probability via bracketing. The interval estimator was achieved by applying different weights to positive and negative classes and training the corresponding weighted large-margin classifiers. They propose to estimate the weighted large-margin classifiers individually. However, empirically the individually estimated classification boundaries may suffer from crossing each other even though, theoretically, they should not.In this work, we propose a technique to ensure non-crossing of the estimated classification boundaries. Furthermore, we take advantage of the estimated conditional class probabilities to precondition our training data. The standard SVM is then applied to the preconditioned training data to achieve robustness. Simulations and real data are used to illustrate their finite sample performance.     

Score tests when a nuisance parameter is unidentified under the null hypothesis

Rao's score test normally replaces nuisance parameters by their maximum likelihood estimates under the null hypothesis about the parameter of interest. In some models, however, a nuisance parameter is not identified under the null, so that this approach cannot be followed. This paper suggests replacing the nuisance parameter by its maximum likelihood estimate from the unrestricted model and making the appropriate adjustment to the variance of the estimated score. This leads to a rather natural modification of Rao's test, which is examined in detail for a regression-type model. It is compared with the approach, which has featured most frequently in the literature on this problem, where a test statistic appropriate to a known value of the nuisance parameter is treated as a function of that parameter and maximised over its range. It is argued that the modified score test has considerable advantages, including robustness to a crucial assumption required by the rival approach.     

Better experimental design by hybridizing binary matching with imbalance optimization

We present a new experimental design procedure that divides a set of experimental units into two groups in order to minimize error in estimating a treatment effect. One concern is the elimination of large covariate imbalance between the two groups before the experiment begins. Another concern is robustness of the design to misspecification in response models. We address both concerns in our proposed design: we first place subjects into pairs using optimal nonbipartite matching, making our estimator robust to complicated nonlinear response models. Our innovation is to keep the matched pairs extant, take differences of the covariate values within each matched pair, and then use the greedy switching heuristic of Krieger et al. (2019) or rerandomization on these differences. This latter step greatly reduces covariate imbalance. Furthermore, our resultant designs are shown to be nearly as random as matching, which is robust to unobserved covariates. When compared to previous designs, our approach exhibits significant improvement in the mean squared error of the treatment effect estimator when the response model is nonlinear and performs at least as well when the response model is linear. Our design procedure can be found as a method in the open source R package available on CRAN called GreedyExperimentalDesign .     

Minimum disparity estimation based on combined disparities: Asymptotic results

The maximum likelihood estimator (MLE) is asymptotically efficient for most parametric models under standard regularity conditions, but it has very poor robustness properties. On the other hand some of the minimum disparity estimators like the minimum Hellinger distance estimator (MHDE) have strong robustness features but their small sample efficiency at the model turns out to be very poor compared to the MLE. Methods based on the minimization of some combined disparities can substantially improve their small sample performances without affecting their robustness properties (Park et al., 1995). All studies involving the combined disparity have so far been empirical, and there are no results on the asymptotic properties of these estimators. In view of the usefulness of these procedures this is a major gap in theory, which we try to fill through the present work. Some illustrations of the performance of the estimators and the corresponding tests are also provided.     

Robust modelling for asset management

We argue that the principal way to achieve robustness is through good elicitation. This means asking the right questions, to elicit only those judgements which can be given reliably. Bayes linear methods minimise the number of prior judgements employed in the analysis. Only first- and second-order moments are specified. We present an example in which a complex problem is modelled in terms of a relatively small number of meaningful prior judgements. Sensitivity to variations in those judgements is explored here and in Goldstein and Wooff (1992).     

Robust and efficient estimation of GARCH models based on Hellinger distance

It is well known that financial data frequently contain outlying observations. Almost all methods and techniques used to estimate GARCH models are likelihood-based and thus generally non-robust against outliers. Minimum distance method, as an important tool for statistical inferences and a competitive alternative for achieving robustness, has surprisingly not been well explored for GARCH models. In this paper, we proposed a minimum Hellinger distance estimator (MHDE) and a minimum profile Hellinger distance estimator (MPHDE), depending on whether the innovation distribution is specified or not, for estimating the parameters in GARCH models. The construction and investigation of the two estimators are quite involved due to the non-i.i.d. nature of data. We proved that the MHDE is a consistent estimator and derived its bias in explicit expression. For both of the proposed estimators, we demonstrated their finite-sample performance through simulation studies and compared with the well-established methods including MLE, Gaussian Quasi-MLE, Non-Gaussian Quasi-MLE and Least Absolute Deviation estimator. Our numerical results showed that MHDE and MPHDE have much better performance than MLE-based methods when data are contaminated while simultaneously they are very competitive when data is clean, which testified to the robustness and efficiency of the two proposed MHD-type estimations.     

Approximations in a group sequential test with unequal group sizes

Our concern in this paper is a group sequential test design for which the sample sizes between interim analyses are not identical. First, we consider a repeated significance test for comparing two treatments in a clinical trial, and study asymptotic properties of the test statistic. Using the arguments developed by Siegmund (1985, Chapters 8 and 9), we then obtain approximations for the overall significance level of the test and for the error level at each interim analysis. Simulation studies are performed to assess the accuracy of the approximations and the robustness of the approximations are examined using numerical examples.     

Robust estimation for longitudinal data under outcome‐dependent visit processes

In longitudinal data where the timing and frequency of the measurement of outcomes may be associated with the value of the outcome, significant bias can occur. Previous results depended on correct specification of the outcome process and a somewhat unrealistic visit process model. In practice, this will never exactly be the case, so it is important to understand to what degree the results hold when those assumptions are violated in order to guide practical use of the methods. This paper presents theory and the results of simulation studies to extend our previous work to more realistic visit process models, as well as Poisson outcomes. We also assess the effects of several types of model misspecification. The estimated bias in these new settings generally mirrors the theoretical and simulation results of our previous work and provides confidence in using maximum likelihood methods in practice. Even when the assumptions about the outcome process did not hold, mixed effects models fit by maximum likelihood produced at most small bias in estimated regression coefficients, illustrating the robustness of these methods. This contrasts with generalised estimating equations approaches where bias increased in the settings of this paper. The analysis of data from a study of change in neurological outcomes following microsurgery for a brain arteriovenous malformation further illustrate the results.     

The Effectiveness of Multistage Ranked Set Sampling in Stratifying the Population

The spectral analysis of Gaussian linear time-series processes is usually based on uni-frequential tools because the spectral density functions of degree 2 and higher are identically zero and there is no polyspectrum in this case. In finite samples, such an approach does not allow the resolution of closely adjacent spectral lines, except by using autoregressive models of excessively high order in the method of maximum entropy. In this article, multi-frequential periodograms designed for the analysis of discrete and mixed spectra are defined and studied for their properties in finite samples. For a given vector of frequencies ω, the sum of squares of the corresponding trigonometric regression model fitted to a time series by unweighted least squares defines the multi-frequential periodogram statistic IM(ω). When ω is unknown, it follows from the properties of nonlinear models whose parameters separate (i.e., the frequencies and the cosine and sine coefficients here) that the least-squares estimator of frequencies is obtained by maximizing I ^M(ω). The first-order, second-order and distribution properties of I ^M(ω) are established theoretically in finite samples, and are compared with those of Schuster's uni-frequential periodogram statistic. In the multi-frequential periodogram analysis, the least-squares estimator of frequencies is proved to be theoretically unbiased in finite samples if the number of periodic components of the time series is correctly estimated. Here, this number is estimated at the end of a stepwise procedure based on pseudo-Flikelihood ratio tests. Simulations are used to compare the stepwise procedure involving I ^M(ω) with a stepwise procedure using Schuster's periodogram, to study an approximation of the asymptotic theory for the frequency estimators in finite samples in relation to the proximity and signal-to-noise ratio of the periodic components, and to assess the robustness of I ^M(ω) against autocorrelation in the analysis of mixed spectra. Overall, the results show an improvement of the new method over the classical approach when spectral lines are adjacent. Finally, three examples with real data illustrate specific aspects of the method, and extensions (i.e., unequally spaced observations, trend modeling, replicated time series, periodogram matrices) are outlined.     

Multi-level models can benefit from minimizing higher-order variations: an illustration using child malnutrition data

ABSTRACT

This study aims to measure the robustness of multi-level models designed for three anthropometric indices – height-for-age (HAZ), weight-for-age (WAZ) and weight-for-height (WHZ) Z-scores for estimating the childhood malnutrition indicators stunting, underweight and wasting in Bangladesh. The 2011 BDHS child malnutrition data have been used in developing multi-level models with and without incorporating specific contextual variables relating to lower administrative units extracted from the 2011 Bangladesh Population and Housing Census. The robustness of the models is examined through (i) testing significance of random effects corresponding to lower administrative units through selection criteria including conditional AIC, R-squared, and LRT; (ii) comparing multi-level model-based estimators to design-based estimators of child malnutrition indicators with their precision at division, district and sub-district levels; and (iii) assessing the impact of contextual variables in capturing higher-order administrative level variations. Findings reveal that the inclusion of important contextual variables helps capture variations at higher-level administrative units, and consequently assists in the selection of robust multi-level models which ultimately provide improved accuracy of estimated parameters. The findings support the application of lower administrative census information in developing a simpler multi-level model by minimizing higher-order variation.     

Monotony relations between distribution functions and their application to experimental design

In a linear regression model an estimator of the unknown coefficients is considered which, in special cases, includes the least squares estimator. In the ease of stable symmetric error distribution and by means of a certain monotony relation between distribution functions optimality of this estimator is proved and the designing problem is investigated. A robustness property of optimal designs against the designing criterion and some conclusions are given concerning the least squares estimator in the case of G- and C-optimality.     

Mean squared error of prediction approach to the analysis of a combined array

SUMMARY The combined array provides a powerful, more statistically rigorous alternative to Taguchi's crossed-array approach to robust parameter design. The combined array assumes a single linear model in the control and the noise factors. One may then find conditions for the control factors which will minimize an appropriate loss function that involves the noise factors. The most appropriate loss function is often simply the resulting process variance, recognizing that the noise factors are actually random effects in the process. Because the major focus of such an experiment is to optimize the estimated process variance, it is vital to understand the resulting prediction properties. This paper develops the mean squared error for the estimated process variance for the combined array approach, under the assumption that the model is correctly specified. Specific combined arrays are compared for robustness. A practical example outlines how this approach may be used to select appropriate combined arrays within a particular experimental situation.     

Empirical Likelihood Ratio Test for a Change-Point in Linear Regression Model

A nonparametric method based on the empirical likelihood is proposed to detect the change-point in the coefficient of linear regression models. The empirical likelihood ratio test statistic is proved to have the same asymptotic null distribution as that with classical parametric likelihood. Under some mild conditions, the maximum empirical likelihood change-point estimator is also shown to be consistent. The simulation results show the sensitivity and robustness of the proposed approach. The method is applied to some real datasets to illustrate the effectiveness.     

Generalized Wald-type tests based on minimum density power divergence estimators

In testing of hypothesis, the robustness of the tests is an important concern. Generally, the maximum likelihood-based tests are most efficient under standard regularity conditions, but they are highly non-robust even under small deviations from the assumed conditions. In this paper, we have proposed generalized Wald-type tests based on minimum density power divergence estimators for parametric hypotheses. This method avoids the use of nonparametric density estimation and the bandwidth selection. The trade-off between efficiency and robustness is controlled by a tuning parameter β. The asymptotic distributions of the test statistics are chi-square with appropriate degrees of freedom. The performance of the proposed tests is explored through simulations and real data analysis.     

Block unimodality for multivariate Bayesian robustness

Summary The development of Bayesian robustness has been growing in the last decade. The theory has extensively dealt with the univariate parameter case. Among the vast amount of proposals in the literature, only a few of them have a straightforward extension to the multivariate case. In this paper we consider the multidimensional version of the class of ε-contaminated prior distributions, with unimodal contaminations. In the multivariate case there is not a unique definition of unimodality and one's choice must be based on statistical ground. Here we propose the use of the block unimodal distributions, which proved to be very suitable for modelling situations where the coordinates of the parameter ϑ are deemed, a priori, weakly correlated.