Note the route to treating problems where one does not have estimating functions differentiable with respect to the parameter of interest. Here the variable in question is discrete. However, by suitably increasing the noise a "resonance" can be set up making the signal apparent. Resonance may be a very important phenomenon scientifically. It has, for example, been proposed as a possible explanation for the ice ages. Consistency and rate of convergence results are available. An information formulation can proceed as follows. Prom the semimartingale representation say, derived from 4.
Asymptotically, G P 2 and Ip are maximized for the same p. But it seems that the periodogram based approach is still appropriate. As a general conclusion, it seems profitable to think about statistical problems in a setting of maximizing an information. I see this as an important unifying principle. Box, G. Davidson, J. Stochastic Limit Theory. Fisher, R. Tests of significance in harmonic analysis.
A , Frieden, B. Physics from Fisher Information. Cambridge U. Press, Cambridge. A Unification, Godambe, V. An optimum property of regular maximum- likelihood estimation. Godambe, V. The foundations of finite sample estimation in stochastic processes. Biometrika 72, Hannan, E. On limit theorems for quadratic functions of discrete time series. Quasi-Likelihood and its Application.
Hutton, J. Quasi-likelihood estimation for semimartingales. Stochastic Process. Kallianpur, G. On the diffusion approximation to a discontinuous model for a single neuron. Sen, Ed. North-Holland, Amsterdam, Kantz, H. Nonlinear Time Series Analysis. Cambridge University Press, Cambridge. Lindsay, B. Springer, New York, to appear. Matthews, R. New Scientist, 30 January , Stochastic integral equations without probability. Bernoulli 6, Norros, L, Valkeila, E.
An elementary approach to a Girsanov formula and other analytical results on fractional Brownian motion. Bernoulli 5, Prakasa Rao, B. Semimartingales and their Statistical Inference. Roussas, G. Contiguity of Probability Measures, Cambridge Univ. Press, London and New York. Thavaneswaran, A. Optimal estimation for semimartingales. Section 2 provides a brief review of this work.
The review is to a large extent based on two papers Barndorff-Nielsen and Shephard a,b where more detailed information may be found. See also Barndorff-Nielsen and Shephard c,d,e. The models in question aim to incorporate one or more of the main stylised features of financial series, be they stock prices, foreign exchange rates or interest rates. A summary of these stylised features, and a comparison with related empirical findings in the study of turbulence, is given in Section 3.
In fact, the intriguing similarities between finance and turbulence have given rise to a new field of study coined 'econophysics'. The study of such processes, as part of probability theory generally, is currently attracting a great deal of attention, see Bertoin , , Sato , Barndorff-Nielsen, Mikosch and Resnick , and references given there. It is by now well recognised that Brownian motion generally provides a poor description of log price processes of stocks and other financial assets.
Merely changing from Brownian motion to another, more suitable, Levy process does not, however, provide a modelling of the important quasi long range dependencies cf. Section 3 that pervades the financial markets. But such dependencies may be captured by further use of Levy processes, as innovation processes driving volatility processes in the framework of SV Stochastic Volatility models. Discrete time models of this kind were considered in Barndorff-Nielsen b. That approach has since been developed, in joint work with Neil Shephard, into the continuous time setting, and the rest of the present note consists mainly in a summary of that work Barndorff-Nielsen and Shephard a,b; cf.
Because of its role in 2. Barndorff-Nielsen b. However, even processes with real long range dependence can be constructed in this way Barndorff-Nielsen Finally, the so-called leverage effect see Section 3 can be modelled by adding an extra term in equation 2. The features are widely recognized as being esssential for understanding and modelling within these two, quite different, subject areas. In finance the observational series concerned consist of values of assets such as stocks or logarithmic stock returns or exchange rates, while in wind turbulence the series typically give the velocities or velocity derivatives or differences , in the mean wind direction of a large Reynolds number wind field.
For some typical examples of empirical probability densities of logarithmic asset returns, on the one hand, and velocity differences in large Reynolds number wind fields, on the other, see, for instance, Eberlein and Keller and Shephard , respectively Barndorff-Nielsen a. A very characteristic trait of time series from turbulence as well as finance is that there seems to be a kind of switching regime between periods of relatively small random fluctuations and periods of high 'activity'.
For detailed and informative discussions of the concepts of intermittency and energy dissipation, see Frisch Stylised features. The generalised hyperbolic laws exhibit this type of behaviour. Velocity differences in turbulence show an inherent asymmetry consistent with Kolmogorov's modified theory of homogeneous high Reynolds number turbulence cf. Barndorff-Nielsen, Distributions of financial asset returns are generally rather close to being symmetric around 0, but for stocks there is a tendency towards asymmetry stemming from the fact that the equity market is prone to react differently to positive as opposed to negative LEVY MODELLING 29 returns, cf.
This reaction pattern, or at least part of it, is referred to as a 'leverage effect' whereby increased volatility tends to be associated with negative returns. By aggregational Gaussianity is meant the fact that long term aggregation of financial asset returns, in the sense of summing the returns over longer periods, will lead to approximately normally distributed variates, and similarly in the turbulence context2. For illustrations of this, see for instance Eberlein and Keller and Barndorff-Nielsen a. The estimated autocorrelation functions based on log price differences on stocks or currencies are generally closely consistent with an assumption of zero autocorrelation.
Nevertheless, this type of financial data exhibit 'quasi long range dependence' which manifests itself inter alia in the empirical autocorrelation functions of the absolute values or the squares of the returns, which stay positive for many lags. For discussions of scaling phenomena in turbulence we refer to Frisch As regards finance, see Barndorff-Nielsen and Prause and references given there. In finance, Burgers' equation has turned up in work by Hodges and Carverhill and Hodges and Selby However, the interpretation of the equation in finance does not appear to have any relation to the role of the equation in turbulence.
References  Barndorff-Nielsen, O. Ada Mechanica 64, In Acccardi, L. Proceedings of a Symposium held October at Columbia University. New York: SpringerVerlag. Finance and Stochastics 2, Theory Prob. Its Appl. To appear. Boston: Birkhauser. Finance and Stochastics, 5, In Barndorff-Nielsen, O. Boston: Birkhauser, Soc, B, 63, Soc, J5, 64, to appear. Cambridge University Press. In Bernard, P. Berlin: Springer.
Bernoulli 1, Finance 9, Economic J. In Dempster, M. Albert-LudwigsUniversitat, Freiburg i. Cox, D. Hinkley and O. Barndorff-Nielsen Eds. London: Chapman and Hall. Asymptotic results are obtained. The process is seen to have a strong local dependence and it's extremal index is computed. A simulation study shows the finite sample size performance of an asymptotic approximation to the distribution of the maximum. Shot noise processes provide a wide class of stochastic models that are particularly well suited to modeling time series with sudden jumps.
Such processes have been applied to modeling river flow data where a rise in the riverflow level could, for example, be attributed to rainfall, Lawrance and Kottegoda and Weiss Moreover, rainfall data, itself, has been modeled via shot noise processes, Waymire and Gupta In applications the sequence of shocks or amplitudes exhibits dependence.
A chain dependent process operates in the following way. Theorectical work on extremes for chain dependent models has been done in Resnick and Neuts , Denzel and O'Brien and more recently in McCormick and Seymour In section 2 we present an extreme value analysis of the shot noise model. Extremes for shot noise processes have been considered by several authors under various assumptions. On the other hand, in practice, the data often appear to contradict such an assumption.
Doney and O'Brien provide an extension to the results of Hsing and Teugels while working under the assumption of constant amplitude. The case of light-tailed amplitudes, viz. Gamma or Weibull distribution, was considered in Homble and McCormick and the heavy-tailed amplitude case, e. Pareto distribution, was developed in McCormick The process under consideration has a strong local dependence quantified by a value referred to as its extremal index.
The method developed in Chernick et al. In section 3 the results of a simulation study are shown. Further, let the interpoint distance d. With the choices made in 2. Hence, letting M! Our first lemma establishes the tail behavior of the stationary distribution for the shot noise process defined in 2. To that end define a variable 2. Assume that the d. Lemma 2. By Corollary 1. Thus condition D un holds. We next turn our attention to computation of the extremal index for the Wk.
Library Hub Discover
The following result is immediate from Lemma 2. Then with un given in 2. Corollary 2. The steps needed to carry out this simulation are specified as follows. The value of L depends stochastically on T, as indicated in the generation of the shot noise process below. Finally, find the maximum of the shot noise process. This procedure is followed times in order to approximate the true underlying distribution. To complete the simulation study, the approximation is computed and then compared to the empirical cumulative distribution function CDF of the simulated process. Each figure is a table of graphs, with the cases indicated by the value of T across the top, and by the impulse response function h x down the left-hand side.
Calculating the extremal index for a class of stationary sequences. Denzel, G. Limit theorems for extreme values of chain-dependent processes.
Doney, R. Loud shot noises. Guttorp, P. Stochastic Modeling of Scientific Data. Chapman and Hall, London. Homble, P. Weak limit results for the extremes of a class of shot noise processes. Prob, 32, Hsing, T. Extremal properties of shot noise processes, Adv. Lawrance, A. Stochastic modeling of riverflow time series. Soc A , McCormick, W. Extremes for shot noise processes with heavy tailed amplitudes. Rates of convergence and approximations to the distribution of the maximum of chain-dependent sequences.
Extremes, 4, Resnick, S. Limit laws for maxima of a sequence of random variables defined on a Markov Chain. Waymire, E. The mathematical structure of rainfall representations 3. Some applications of the point process theory to rainfall processes. Water Resources Res. Weiss, G. Filtered Poisson processes as models for daily streamflow data. PhD thesis. Imperial College, London. Prakasa Rao Indian Statistical Institute, New Delhi Abstract Stochastic partial differential equations SPDE are used for stochastic modelling , for instance, in the study of neuronal behaviour in neurophysiology, in modelling sea surface temparature and sea surface height in physical oceanography , and in building stochastic models for turbulence.
Probabilistic theory underlying the subject of SPDE is discussed in Ito and more recently in Kallianpur and Xiong among others. The study of statistical inference for the parameters involved in SPDE is more recent. Asymptotic theory of maximum likelihood estimators for a class of SPDE is discussed in Huebner, Khasminskii and Rozovskii and Huebner and Rozovskii following methods in Ibragimov and Khasminskii Bayes estimation problems for such a class of SPDE are investigated in Prakasa Rao , following the techniques developed in Borwanker et al.
An analogue of the Bernstein-von Mises theorem for parabolic stochastic partial differential equations is proved in Prakasa Rao As a consequence, the asymptotic properties of the Bayes estimators of the parameters are investigated using the asymptotic properties of maximum likelihood estimators proved in Huebner and Rozovskii Nonparametric estimation of a linear multiplier for some classes of SPDE are studied in Prakasa Rao a,b by the kernel method of density estimation following the techniques in Kutoyants Kallianpur and Xiong Huebner et al.
Statistical inference for diffusion type processes and semimartingales in general is studied in Prakasa Rao a,b. Our aim in this paper is to review some recent work of us on the Bernstein-von Mises type theorems for parabolic SPDE and to present some new results on the problem of estimation of a linear multiplier for a class of SPDE using the methods of nonparametric inference following the approach of Kutoyants For the definition of cylindrical Brownian motion, see, Kallianpur and Xiong , p. The order Ord A of a partial differential operator A is defined to be the order of the highest partial derivative in A.
We follow the notation introduced in Huebner and Rozovskii Assume that the following conditions hold. We assume that the following condition holds. We identify the spaces HQ denoted by Hs and the norms! The conditons H1 - H5 described above are the same as those in Huebner and Rozovskii Let dP? We now have the following main theorem which is an analogue of the Bernstein-von Mises theorem cf.
Operator-Related Function Theory and Time-Frequency Analysis
Prakasa Rao , for diffusion processes and diffusion fields. Theorem 2. Then OO lim 1. Suppose the condition Dl holds. As a particular case of Theorem 2. For proofs of Theorems 2. We omit the proof. Theorem 3. Then Oa. As a consequence of Theorem 3. Remarks: A general approach for the study of asymptotic properties of maximum likelihood estimators and Bayes estimators is by proving the local asymptotic normality of the loglikelihood ratio process as was done in 58 PRAKASA RAO Prakasa Rao , Ibragimov and Khasminskii in the classical i.
Our approach for Bayes estimation, via the comparison of the rates of convergence of the difference between the maximum likelihood estimator and the Bayes estimator, is a consequence of the the Bernstein - Von Mises type theorem. We now consider a nonparametric version of the problem discussed earlier for a class of SPDE. Rozovskii Further suppose that the kernel G. The third term is bounded by where C2 is an absolute constant. Relations 4. Kutoyants , p. Theorem 4. One can study the problem of nonparametric inference for a linear multiplier for such a class of SPDE by the above methods cf.
Prakasa Rao b. References Borwanker, J. Math Statist. Huebner, M. Cambanis, J. Ghosh, R. Karandikar, P. Sen, Springer, New York, pp. On asymptotic properties of maximum likelihood estimators for parabolic stochastic SPDE's. Theory and Relat Fields, , Ibragimov, LA. Ito, K. Kutoyants, Yu. Estimation of the location of the cusp of a continuous density, Ann. The Bernstein - von Mises theorem for a class of diffusion processes, Teor. On Bayes estimation for diffusion fields. Ghosh and J.
Roy, Statistical Publishing Society, Calcutta. Bayes estimation for parabolic stochastic partial differential equations, Preprint, Indian Statistical Institute, New Delhi. Bayes estimation for some stochastic partial differential equations , J. Nonparametric inference for a class of stochastic partial differential equations, Tech. Report No. Nonparametric inference for a class of stochastic partial differential equations II, Statist. Infer, for Stock. Rozovskii, B. Stochastic Evolution Systems, Kluwer, Dordrecht.
Shimakura, N. Soc, Providence. Key words and phrases: Fixed design regression, stationarity, weights, fixed design regression estimate, asymptotic unbiasedness, consistency in quadratic mean, association, asymptotic normality. All results in this paper hold for all x G 3id and with xn as defined above.
Since the first term on the right-hand side of 2. At this point, it is to be observed that Theorems 2. The property of association is used only in Theorem 3. Although some of the assumptions spelled out below coincide with assumptions previously made, we choose to gather all of them here for easy reference. Under assumptions Cl - C5 , the convergence asserted in 3. The proof of the theorem follows by combining the two propositions below. The propositions, as well as the three lemmas employed in this section, hold under all or parts only of assumptions Cl - C5.
However, these lesser assumptions will not be explicitly stated. Proposition 3. The convergence asserted in 3. Assuming for a moment that Propositions 3. It follows from Propositions 3. Let ynm and y nm be defined by 3. Corollary 3. Proof, i The right-hand side of the expression in Lemma 3. However, this last expression tends to 0 by assumption C2. Proof of Proposition 3. Relations 3. The proof of the proposition is completed. Then the r. Proof, i From 3. But by Proposition 3. Then relation 3. The proof of the lemma is completed. With the ynms denned by 3. Lemma 3. This fact along with 3.
The author is indebted to the referees and one of the editors whose thoughtful comments led to the removal of an inconsistency in the assumptions and helped improve a previous version of this paper. Limit theorems for negatively dependent random variables. Bulinski, A. On the convergence rates in the CLT for positively and negatively dependent random fields.
Ibragimov and A. Zaitsev, Eds , pp. Gordon and Breach Publishers, Amsterdam. Cai, Z. Weak convergence for smooth estimator of a distribution function under negative association. Stochastic Analysis and Applications 17, - Berry-Essen bounds for a smooth estimator of a distribution function under association. Journal of Nonparametric Statistics 11, 79 - Gasser, T. Kernel estimation of regression function. Gasser and M. Rosenblatt, Eds. Lecture Notes in Mathematics, Vol. Springer-Verlag, Berlin - New York. Georgiev, A. Consistent nonparametric multiple regression: The fixed design case.
Nonparametric function recovering from noisy observations. Journal of Statistical Planning and Inference 13, 1 - Loeve, M. Probability Theory, 3rd ed. Van Nostrand, Princeton, NJ. Newman, C M. Normal fluctuations and the FKG inequalities. Communications in Mathematical Physics 75, - Patterson, R. Strong laws of large numbers for negatively dependent random elements.
Nonlinear Analysis, Theory, Methods and Applications 30, - Priestley, M. Nonparametric function fitting. Journal of the Royal Statistical Society Ser. B 34, - Consistent regression estimation with fixed design points under dependent conditions. Statistics and Probability Letters 8, 41 - Fixed design regression for time series: Asymptotic normality.
Journal of Multivariate Analysis 40, - Asymptotic normality of random fields of positively or negatively associated processes. Journal of Multivariate Analysis 50, - Positive and negative dependence with some statistical applications. Ghosh, Ed. Marcell Dekker, Inc. Asymptotic normality of the kernel estimate of a probability density function.
Statistics and Probability Letters 50, Taylor, R. Negative dependence in Banach spaces and laws of large numbers. A strong law of large numbers for arrays of rowwise negatively dependent random variables.
- The Nature of God: An Inquiry into Divine Attributes (Cornell Studies in the Philosophy of Religion)!
- Akaike information criterion.
- About these proceedings;
- Akaike information criterion - Wikipedia.
Also, to appear in Journal of Stochastic Analysis and Applications. Weak laws of large numbers for arrays of rowwise negatively dependent random variables. Taylor Department of Statistics The University of Georgia Athens, GA Abstract A dependent bootstrap is shown to produce estimators which have smaller variances but which are still consistent and asymptotically valid.
Simulated confidence intervals are used to examine possible gains in coverage probabilities and interval lengths. The use of independent models is prevalent in bootstrapping. Efron introduced the bootstrap as a tool to estimate the standard error of a statistic, and an enormous amount of applied and theoretical research on the bootstrap technique has followed in the past two decades. While much of this ensuing research has been methodological adaptions and theoretical validity verifications for different statistics, considerable research has been directed toward shortcomings and possible improvements to the basic bootstrap technique.
The traditional resampling of the sample observations with replacement produces independent and identically distributed bootstrap random variables conditional on the original sample , and many of the theoretical justifications of the bootstrap procedures are crucially related to techniques involving independent random variables. Resampling without replacement produces dependent random variables actually negatively dependent which are still identically distributed and in fact has the desirable property of exchangeability. The purpose of this paper is to consider some estimation using a form of dependent bootstrapping.
In particular, confidence interval comparisons will be given for the traditional bootstrap procedure and the dependent bootstrap procedure. The majority of research on resampling without replacement has been for application in finite population sampling. Gross introduced the concept and many others Bickel and Freedman, ; Chao and Lo, ; Sitter, ; Booth,Butler and Hall, ; and others have extended this research. In Politis and Romano examined resampling without replacement from a data set to approximate the sampling distribution of a statistic Tn.
Under weak assumptions, they showed that the empirical distribution of the suitably normalized values of the statistic computed for all subsamples of size b from the original data is first order asymptotically valid for the true sampling distribution of Tn. This is a generalization from Wu who studied the same method in the i. Bertail showed second order correctness of this method for an adequately chosen resample size.
Their investigations differ from this proposed research because they sample without replacement from the original data rather than an enriched collection with a fixed number of copies of each observation. Praestgaard and Wellner showed that "ra out of fcn" could allow larger bootstrap sample sizes and some asymptotic results using exchangeability arguments. Babu and Singh and Babu and Bai showed that Edgeworth expansions could be used to obtain approximation results for estimators based on samples drawn without replacement from a finite population.
Their approximation results provide for weak convergence of normalized absolute differences of original sample statistics and bootstrap statistics. This paper will compare the coverage probabilities and lengths of the more generally used bootstrap confidence intervals for the traditional bootstrap and the dependent bootstrap. The formal definition of the dependent bootstrap procedure and the theoretical properties of consistency and asymptotic validity are listed in Section 2. Section 3 provides the description and results of the simulations for the confidence intervals.
Negative dependence includes independence, and the terminology of negative relates to 2. In particular, they showed that the dependent bootstrap produces negatively dependent random variables. The technique of the proof for Theorem 2. Moreover, it is important to observe that required moment conditions in Theorem 2. Finite population versions of these results were also obtained cf: Smith and Taylor For the validity asymptotic normality for the dependent bootstrap a stronger more restrictive form of negative dependence is needed, namely negative association.
Using a combination of exchangeable and negative association results, Patterson, Smith, Taylor, and Bozorgnia obtained the following central limit thoerem for the dependent bootstrap. Moreover, Theorem 2. This result is stated for i. Using the bootstrap samples, the percentile and bootstrap-t confidence intervals were obtained. The same procedure was followed using the traditional bootstrap. The estimated coverage probabilities and average lengths are reported in Tables The results show that for all distributions there is little difference between the coverage probabilities and the lengths of the normal theory CI, the traditional bootstrap and the dependent bootstrap procedure, when the bootstrap-t method of CI computation is used.
However, for the percentile method of confidence interval computation, the dependent bootstrap performs poorly when compared to the traditional bootstrap and the normal theory method cf: Tables This result is to be expected since the variance of the dependent bootstrap mean estimator cf: 2. Thus, an adjustment to the percentile method is needed for the dependent bootstrap confidence intervals to achieve desired coverage probabilities while trying to maintain shorter lengths.
Specifically, since the percentiles are functionally related to the standard deviation, instead of using 3. Laws of large numbers for bootstrapped U-statistics. Planning and Inf. Mixtures of global and local Edgeworth expansions and their applications. Multivariate Anal, 59, Babu, G. Edgeworth expansions for sampling without replacement from finite populations.
Multivariate Anal, 17, Bertail, P. Second-order properties of an extrapolated bootstrap without replacement under weak assumptions. Bernoulli, 3, Bickel, P. Resampling fewer than n observations: gains, losses, and remedies for losses. Statistica Sinica, 7, Asymptotic normality and the bootstrap in stratified sampling. Stat, 12, Booth, J. Bootstrap methods for finite populations.
Link to external resource:
Assoc, 89, Bozorgnia, A. Chung type strong laws for arrays of random elements and bootstrapping. Anal, and AppL, 15, Chao, M. A bootstrap method for finite populations. Sankaya, Series A. DiCiccio, T. Bootstrap confidence intervals. Scl, 11, Efron, B. Bootstrap methods: Another look at the jackknife. Stat, 7, Gross, S. Median estimation in sample surveys. Survey Res. Methods, Amer. Assoc, Hu, T. On the strong law for arrays and the bootstrap mean and variance. Math, and Math. NonLinear Analysis, 47, Politis, D. Large sample confidence regions based on subsamples under minimal assumptions.
Praestgaard, J. Exchangeably weighted bootstraps of the general empirical process. Singh, K. Breakdown theory for bootstrap quantiles. Sitter, R. Comparing three bootstrap methods for survey data. Stat, 20, Smith, W. Math, and Management Sciences, Wu, C. On the asymptotic properties of the jackknife historgram. Stat, 18, Traditional Bootstrap 20 40 2. These tests are based on autoregression rank scores, and extend to the time-series context a method proposed by Jureckova for regression rank scores and regression models with independent observations.
Their asymptotic distributions are derived, and they are shown to coincide with those of classical Kolmogorov-Smirnov statistics, under null hypotheses as well as under contiguous alternatives. Local asymptotic efficiencies are investigated. A Monte Carlo experiment is carried out to illustrate the performance of the proposed tests. Keywords: Autoregressive models, Autoregression quantiles, Autoregression rank scores, Kolmogorov-Smirnov test, Local asymptotic efficiency.
Koul and Saleh , extending to the time-series context Koenker and Bassett 's concept of regression quantiles, defined the a-autoregression quantile for model 1. A crucial property of autoregression rank scores is their autoregressioninvariance, i. Some further algebraic relations between autoregression quantiles and the corresponding autoregression rank scores are provided in Lemma 2.
In the present paper, new tests based on a autoregression rank score version of the traditional Kolmogorov-Smirnov statistic are introduced for model 1. The asymptotic behaviour of these tests is investigated in Section 3, where we show that the limiting distributions of the test statistics coincide with those of the classical Kolmogorov-Smirnov statistics, both under the null hypothesis as under contiguous alternatives.
Our results extend those of Jureckova from regression models to autoregression models. The local asymptotic efficiency of these tests is also investigated. Finally, the performance of the proposed tests is illustrated on simulated AR series with Normal, Laplace and Cauchy innovation densities, respectively.
STOC Proceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing
Such tests play a crucial role, for instance, in the order identification process see Garel and Hallin D Proof Theorem 2. Prom Theorem V. This proof, as well as the proof of Theorem 3. As for the linear approximation 2. This problem already appears in the classical case of Kolmogorov-Smirnov tests for linear models with independent observations; see Hajek and Sidak , Section VII. However, local ARE results can be obtained from the linear approximation 2. If we apply the Cauchy-Schwarz inequality to 3. Though more extensive simulations should be undertaken, Table 1 nevertheless indicates that Kolmogorov-Smirnov techniques, especially the onesided, based on nonlinear test statistics, can be expected to beat locally asymptotically optimal tests based on linear statistics.
The authors thank the Editor, Professor Ishwar Basawa, and an anonymous referee for their careful reading of the manuscript and helpful comments. References  Garel, B. Hallin Rank-based AR order identification. Journal of the American Statistical Association 94, Jureckova Regression rank scores and regression quantiles. Annals of Statistics 20, Jureckova, R. Koenker, and S. Portnoy Tests of linear hypotheses based on regression rank scores. Journal of Nonparametric Statistics 2, Extension of the Kolmogorov-Smirnov test to the regression alternatives.
Le Cam and J. Neyman, eds , Berlin, Springer-Verlag. Sidak Theory of Rank Tests. Sidak, and P. Sen Theory of Rank Tests 2nd edition. Jureckova, J. Kalvova, J. Picek, and T. Zahaf Nonparametric tests in AR models, with applications to climatic data. Environmetrics 8, Nonparametric tests of independence of two autoregressive time series based on autoregression rank scores. Journal of Statistical Planning and Inference 75, Optimal tests for autoregression models based on autoregression rank scores. Annals of Statistics 27, Puri Aligned rank tests for linear models with autocorrelated errors.
Journal of Multivariate Analysis 50, Werker Optimal testing for semiparametric autoregressive models : from Gaussian Lagrange multipliers to regression rank scores and adaptive tests. Dekker, New York. Autoregression quantiles and related rank score processes for generalized random coefficient autoregressive processes. Journal of Statistical Planning and Inference 68, Tests of Kolmogorov-Smirnov type based on regression rank scores. Academia, Prague. Regression rank scores against heavy-tailed alternatives. Bassett Regression quantiles.
Econometrica 46, Algorithm AS : Computing regression quantiles. Remark AS R A remark on Algorithm AS : Computing dual regression quantiles and regression rank scores. It is expected that most of the accepted papers will appear subsequently, possibly in a more complete form, in peer-reviewed scientific journals. The program committee thanks all the authors for supporting the conference through their submissions.
Last but not least, I would like to thank the program committee members for their expertise, dedication and effort. Special thanks go to David Kempe, who served de-facto as a co-chair at several stages of the decision process. Sign in Help View Cart. Manage this Book. Add to my favorites. Recommend to Library. Email to a friend. Digg This. Notify Me! E-mail Alerts. RSS Feeds. Title Information. Series: Proceedings. Editor s : Robert Krauthgamer. The latter paper was also given the Best Student Paper award. Return to All Sections. Front Matter. Locality-sensitive Hashing without False Negatives.
New directions in nearest neighbor searching with applications to lattice sieving. Phase Transitions in Group Testing. Designing Networks with Good Equilibria under Uncertainty. Characterisation of Strongly Stable Matchings. Learning and Efficiency in Games with Dynamic Population. The Complexity of All-switches Strategy Improvement. Make-to-Order Integrated Scheduling and Distribution. Online Pricing with Impatient Bidders. Local-on-Average Distributed Tasks. Expanders via Local Edge Flips. Nearly-optimal bounds for sparse recovery in generic norms, with applications to k -median sketching.
Simpler, faster and shorter labels for distances in graphs. Deterministic Algorithms for Submodular Maximization Problems.
- SESSION: New techniques and environments!
- Statistics Canada International Symposium Series: Proceedings.
- Recommended for you!
- Proceedings of the eighth annual forest inventory and analysis symposium | Treesearch.
- The Handbook of Language and Globalization?
- The Thirteen Books of the Elements, Vol. 3: Books 10-13.
- Numerical Simulation of Plasmas;
The complexity of approximately counting in 2-spin systems on k -uniform bounded-degree hypergraphs. Sparse Approximation via Generating Point Sets. Jointly Private Convex Programming. Time vs. Markovian Hitters and the Complexity of Blind Rendezvous. Stabilizing Consensus with Many Opinions.
Balanced Allocation: Patience is not a Virtue. Weighted dynamic finger in binary search trees. Approximating Low-Stretch Spanners. Better Distance Preservers and Additive Spanners. Near-Optimal Light Spanners. Efficient quantum algorithms for computing class groups and solving the principal ideal problem in arbitrary degree number fields. Permutation patterns are hard to count. Computing in continuous space with self-assembling polygonal tiles extended abstract.
Beyond the Richter-Thomassen Conjecture. Treetopes and their Graphs. Online Contention Resolution Schemes. Finding a Stable Allocation in Polymatroid Intersection. The matching problem has no small symmetric SDP. Linear Recognition of Almost Interval Graphs. Subexponential parameterized algorithm for Interval Completion. Lower bounds for the parameterized complexity of Minimum Fill-In and other completion problems. Directed multicut is W -hard, even for four terminal pairs.
Persistent Homology and Nested Dissection. Subtree Isomorphism Revisited. Algorithmic Complexity of Power Law Networks. Clustering Problems on Sliding Windows. On the Complexity of Dynamic Mechanism Design. Packing Small Vectors. Approximation schemes for machine scheduling with resource in- dependent processing times. Improved Approximation for Vector Bin Packing. Connectivity in bridge-addable graph classes: the McDiarmid-Steger-Welsh conjecture. Tight conditional lower bounds for counting perfect matchings on graphs of bounded treewidth, cliquewidth, and genus. Blocking Optimal k -Arborescences.
Constructive algorithm for path-width of matroids. Approximation of non-boolean 2CSP. Algorithms and Adaptivity Gaps for Stochastic Probing.