-
作者：Janson, Lucas; Fithian, William; Hastie, Trevor J.
作者单位：Stanford University
摘要：To most applied statisticians, a fitting procedure's degrees of freedom is synonymous with its model complexity, or its capacity for overfitting to data. In particular, the degrees of freedom is often used to parameterize the bias-variance trade-off in model selection. We argue that, on the contrary, model complexity and degrees of freedom may correspond very poorly. We exhibit and theoretically explore various fitting procedures for which the degrees of freedom is not monotonic in the model c...
-
作者：Doucet, A.; Pitt, M. K.; Deligiannidis, G.; Kohn, R.
作者单位：University of Oxford; University of Warwick; University of New South Wales Sydney
摘要：When an unbiased estimator of the likelihood is used within a Metropolis-Hastings chain, it is necessary to trade off the number of Monte Carlo samples used to construct this estimator against the asymptotic variances of the averages computed under this chain. Using many Monte Carlo samples will typically result in Metropolis-Hastings averages with lower asymptotic variances than the corresponding averages that use fewer samples; however, the computing time required to construct the likelihood...
-
作者：Kato, Shogo; Jones, M. C.
作者单位：Research Organization of Information & Systems (ROIS); Institute of Statistical Mathematics (ISM) - Japan; Open University - UK
摘要：This article presents a class of four-parameter distributions for circular data that are unimodal, possess simple characteristic and density functions and a tractable distribution function, can be interpretably parameterized directly in terms of their trigonometric moments, afford a very wide range of skewness and kurtosis, envelop numerous interesting submodels including the wrapped Cauchy and cardioid distributions, allow straightforward parameter estimation by both method of moments and max...
-
作者：Konstantinou, M.; Dette, H.
作者单位：Ruhr University Bochum
摘要：We consider the construction of optimal designs for nonlinear regression models when there are measurement errors in the covariates. Corresponding approximate design theory is developed for maximum likelihood and least-squares estimation, with the latter leading to nonconcave optimization problems. Analytical characterizations of the locally D-optimal saturated designs are provided for the Michaelis-Menten, E-max and exponential regression models. Through concrete applications, we illustrate h...
-
作者：Ma, Yanyuan; Zhang, Xinyu
作者单位：University of South Carolina System; University of South Carolina Columbia; Capital University of Economics & Business
摘要：A crucial component of performing sufficient dimension reduction is to determine the structural dimension of the reduction model. We propose a novel information criterion-based method for this purpose, a special feature of which is that when examining the goodness-of-fit of the current model, one needs to perform model evaluation by using an enlarged candidate model. Although the procedure does not require estimation under the enlarged model of dimension k + 1, the decision as to how well the ...
-
作者：Chakraborty, Anirvan; Chaudhuri, Probal
作者单位：Indian Statistical Institute; Indian Statistical Institute Kolkata
摘要：The Wilcoxon-Mann-Whitney test is a robust competitor of the test in the univariate setting. For finite-dimensional multivariate non-Gaussian data, several extensions of the Wilcoxon-Mann-Whitney test have been shown to outperform Hotelling's test. In this paper, we study a Wilcoxon-Mann-Whitney-type test based on spatial ranks in infinite-dimensional spaces, we investigate its asymptotic properties and compare it with several existing tests. The proposed test is shown to be robust with respec...
-
作者：Wu, Yuanshan; Yin, Guosheng
作者单位：Wuhan University; University of Hong Kong
摘要：To accommodate the heterogeneity that is often present in ultrahigh-dimensional data, we propose a conditional quantile screening method, which enables us to select features that contribute to the conditional quantile of the response given the covariates. The method can naturally handle censored data by incorporating a weighting scheme through redistribution of the mass to the right; moreover, it is invariant to monotone transformation of the response and requires substantially weaker conditio...
-
作者：Rootzen, Holger; Zholud, Dmitrii
作者单位：Chalmers University of Technology; University of Gothenburg
摘要：This paper develops tail estimation methods to handle false positives in multiple testing problems where testing is done at extreme significance levels and with low degrees of freedom, and where the true null distribution may differ from the theoretical one. We show that the number of false positives, conditional on the total number of positives, has an approximately binomial distribution, and we find estimators of the distribution parameter. We also develop methods for estimation of the true ...
-
作者：Scott, J. G.; Shively, T. S.; Walker, S. G.
作者单位：University of Texas System; University of Texas Austin; University of Texas System; University of Texas Austin
摘要：This paper adopts a nonparametric Bayesian approach to testing whether a function is monotone. Two new families of tests are constructed. The first uses constrained smoothing splines with a hierarchical stochastic-process prior that explicitly controls the prior probability of monotonicity. The second uses regression splines together with two proposals for the prior over the regression coefficients. Via simulation, the finite-sample performance of the tests is shown to improve upon existing fr...
-
作者：Cao, Hongyuan; Wu, Wei Biao
作者单位：University of Missouri System; University of Missouri Columbia; University of Chicago
摘要：We consider large scale multiple testing for data that have locally clustered signals. With this structure, we apply techniques from changepoint analysis and propose a boundary detection algorithm so that the clustering information can be utilized. Consequently the precision of the multiple testing procedure is substantially improved. We study tests with independent as well as dependent p-values. Monte Carlo simulations suggest that the methods perform well with realistic sample sizes and show...
