Showing papers in "Journal of Statistical Computation and Simulation in 1984"
70 citations
TL;DR: In this article, a multivariate central limit theorem is established for such variables in a broad class of regenerative queueing systems, and robust, asymptotically stable variance reduction techniques that incorporate these concomitant variables into poststratified sampling schemes as well as control-variate schemes.
Abstract: To improve the efficiency of system performance estimators generated by a queueing simulation, procedures are developed for exploiting standardized concomitant variables that are associated with each input process sampled during the simulation. A multivariate central limit theorem is established for such variables in a broad class of regenerative queueing systems. This result is the basis for robust, asymptotically stable variance reduction techniques that incorporate these concomitant variables into poststratified sampling schemes as well as control-variate schemes. Each procedure is adapted to estimation methods based on replication analysis and regenerative analysis. A summary of the results of an experimental performance evaluation indicates the potential efficiency gains that can be achieved with these procedures
37 citations
TL;DR: In this paper, a numerical method for solving the renewal equation is proposed, which generates a cubic spline approximation of the renewal function by the Galerkin technique, tested on Gamma lifetime densities of various shapes.
Abstract: A numerical method for solving the renewal equation is proposed. The method which generates a cubic spline approximation of the renewal function by the Galerkin technique is tested on Gamma lifetime densities of various shapes. Results are compared against known analytical solutions and earlier approximation.
32 citations
TL;DR: The multivariate response simulation model is defined and some historical solutions from the literature are discussed, including the statistical rationale for the use of these techniques, the advantage of using them in simulation output analysis, and a comprehensive survey of the simulation literature which has applied these procedures to date.
Abstract: The multiple response problem in simulation analysis refers to the statistical design and analysis of simulation experiments which output more than a single response variable, or measure of effectiveness. In this article, the multivariate response simulation model is defined and some historical solutions from the literature are discussed. Recent research into the use of multivariate statistical methods in simulation analysis is reviewed, including the statistical rationale for the use of these techniques, the advantage of using these methods in simulation output analysis, and a comprehensive survey of the simulation literature which has applied these procedures to date. The concept of multivariate response simulation metamodel, an auxiliary, analytic model which serves to aid in the interpretation of the simulation model, is also presented.
24 citations
TL;DR: In this article, two methods for estimating the variance function Var{Y\X = X} in weighted regression analysis are studied via simulation, and the results show that the first method can perform very poorly in the presence of strong heteroscedasticity, and that the second method may provide an effective alternative in such cases.
Abstract: Two methods for estimating the variance function Var{Y\X = X) in weighted regression analysis are studied via simulation.The first method assumes that Var(Y|X= x) is a known power of the regression function E(Y|X= x), and estimates the coefficient vector susing iteratively reweighted least squares. In the second method, Var(Y|X= x) is estimated directly using a nonparametric regression technique.The results show that the first method can perform very poorly in the presence of strong heteroscedasticity, and that the second method may provide an effective alternative in such cases.
16 citations
TL;DR: In this paper, the effect of multiple testing on type I error was investigated by simulating two-sample data and studying five common tests: the t, Wilcoxon-Mann-Whitney,t on logs, Yuen-Dixon trimmed t, and Welch's test.
Abstract: In many situations a variety of tests are available to test essentially the same null hypothesis. In practice the statistician who fails to reject with the first test used will sometimes try several others, stopping when he obtains the hoped-for significance. This raises the type I error rate, but no broad study has previously been made to address the question by how much. Here the effect of such multiple testing is investigated by simulating two-sample data and studying five common tests: the t, Wilcoxon-Mann-Whitney,t on logs, Yuen-Dixon trimmed t, and Welch's test.
14 citations
TL;DR: In this paper, a multiple range subset selection procedure for k normal populations with a common known variance is proposed, and Monte Carlo methods are used to compare it with the multiple F subsets selection procedure of Somerville (1983) and with the method of Gupta (1965)
Abstract: A multiple range subset selection procedure for k normal populations with a common known variance is proposed. Monte Carlo methods are used to compare it with the multiple F subset selection procedure of Somerville (1983) and with the method of Gupta (1965)
14 citations
TL;DR: In this paper, a class of prior distributions is defined to reflect exchangeability of a set of binomial probabilities, indexed by the hyperparameter K, which indicates the precision of the user's prior belief about the similarity of the probabilities.
Abstract: A class of prior distributions is defined to reflect exchangeability of a set of binomial probabilities. The class is indexed by the hyperparameter K, which indicates the precision of the user's prior belief about the similarity of the probabilities. By estimating the unknown value of K from the marginal distribution, simple new point and interval estimators are proposed.
13 citations
TL;DR: In this paper, the authors provided an empirical study of the behavior of the X 2 test under H 0 as well as under a variety of alternatives, and found that the Pearson test was the most powerful among the tests which do not inflate the nominal significance level.
Abstract: Several tests are available for the hypothesis stating that there is no 3-factor interaction in a 2 × 2 × 2 contingency table. This paper provides an empirical study of the behavior, under H 0 well as under a variety of alternatives, of six tests. Pearson's X 2 test, which was first applied to the 2 × 2 × 2 table by Bartlett (1935) is found most powerful among the tests which do not inflate the nominal significance level.
12 citations
TL;DR: In this article, a class of estimators is proposed for the difference of means when sampling from a biv ariate normal distribution with means w, and u2, variances o\ and
Abstract: A class of estimators is proposed for the difference of means when sampling from a biv ariate normal distribution with means w, and u2, variances o\ and
12 citations
TL;DR: In this paper, simple and exact methods for computer generation of bivariate Beta random variables are considered and comparative timings over a range of values of the distribution parameters are given and some recommendations are made.
Abstract: Various simple and exact methods for computer generation of bivariate Beta random variables are considered. Comparative timings over a range of values of the distribution parameters are given and some recommendations are made.
TL;DR: In this paper, a class of goodness-of-fit test statistics which are calculated directly from probability plots as they are constructed in practice is described, and empirical sampling methods are used to derive the null distribution of each of the corresponding test statistics.
Abstract: Probability plots are commonly used as a technique for testing distributional assumptions. However, any conclusion about the linearity of such a plot is based strictly on the user's judgment. Regression tests of fit are supposed to make this procedure more objective, but these tests typically are not based on probability plots as they are constructed in practice. This is because the developers of these tests defined probability plots in terms of plotting positions which are not used by practitioners. In this paper, a class of goodness-of-fit test statistics which are calculated directly from probability plots as they are constructed in practice is described. Several realistic plotting positions for the normal distribution are chosen and empirical sampling methods are used to derive the null distribution of each of the corresponding test statistics. These tests are then compared on the basis of 5% power against certain nonnormal alternatives. Results of the comparisons indicate that the test based on the p...
TL;DR: In this article, the authors evaluate the distribution function of a bivariate gamma distribution via a power series expansion in the correlation coefficient, which may be performed via a Power-Series Expansion in the Correlation coefficient.
Abstract: In this paper, we evaluate the distribution function of a bivariate gamma distribution which arises frequently in the literature.The calculation may be performed via a power series expansion in the correlation coefficient.
TL;DR: It is shown by example how various well-known probability inequalities can aid in the design of general nonuniform random variate generators, i.e. generators that can be used for large classes of densities.
Abstract: We show by example how various well-known probability inequalities can aid in the design of general nonuniform random variate generators, i.e. generators that can be used for large classes of densities. We obtain universal algorithms for the following classes:(1)all unimodal densities with mode at 0 and absolute rth moment not exceeding a known constant;(2)all densities satisfying a Lipschitz condition with known constant, having an absolute rth moment or a moment generating function in a neighborhood of 0 not exceeding a known constant. The algorithms assume that the density f can be computed at every point, but the distribution function is not needed.
TL;DR: In this paper, the estimation of a collection of location parameters when it is believed, a priori, that their ordering is known is considered and robust estimators which perform well for a broad range of distributions are discussed.
Abstract: We consider the estimation of a collection of location parameters when it is believed, a priori, that their ordering is known. The least squares and least absolute deviations estimates subject to this ordering restriction have been studied in the literature. We seek robust estimators which perform well for a broad range of distributions. The results of a Monte Carlo study and a study of computation algorithms are discussed.
TL;DR: A sharpening of the harmonic-mean rule of thumb for combining tests in parallel was proposed in this paper, where the harmonic mean rule was replaced with a rule of combining tests "in parallel".
Abstract: (1984). C213. A sharpening of the harmonic-mean rule of thumb for combining tests ”in parallel“. Journal of Statistical Computation and Simulation: Vol. 20, No. 2, pp. 173-176.
TL;DR: In this article, the problem of putting confidence bands on a logistic cumulative distribution function is considered, and confidence bands are constructed using Kolmogorov-Smirnov-type statistics.
Abstract: The problem of putting confidence bands on a logistic cumulative distribution function is considered. One and two-sided confidence bands on the entire cumulative distribution function and simultaneous confidence intervals for the interval probabilities under the distribution are constructed using Kolmogorov—Smirnov-type statistics. Small sample and asymptotic distributions of the relevant statistics arc provided so that the construction can be carried out in any practical situation.
TL;DR: In this paper, the impact of random number generators in Monte Carlo time series simulation is discussed, using six different random number generator types and a large number of autoregressive time series of order one are generated.
Abstract: In this paper the impact of random number generators in Monte Carlo time series simulation is discussed. Using six different random number generators a large number of autoregressive time series of order one are generated and some important statistics are compared. From this it seems that unless a very poor generator is used, the choice of generator has little or no impact on the conclusions drawn on the basis of a simulation study.
TL;DR: In this paper, the effect of nonnegativity constraint on mean squared error is compared to the alternative of replacing negative values of unconstrained estimates to zero, for both the REML and ML approaches.
Abstract: Two variations of the dispersion-mean correspondence model for varince components lead to the ML and REML equations. This formulation provides for addition of a nonnegativity constraint to the computational method of T. W. Anderson (1971, 1973), an iterative procedure for obtaining ML and REML estimates, which not only assures that estimates will be within the parameter space, but contributes to the stability of the algorithm as well. Estimates of the variance components obtained in this way (which includes MINQUE) are simulated for four balanced and two unbalanced random effects designs. The effect of the nonnegativity constraint on mean squared error is compared to the alternative of replacing negative values of unconstrained estimates to zero, for both the REML and ML approaches. For the two unbalanced designs, the effect of initial starting values and the significance of iterating are also assessed. The simulation results are an extension of those reported by Rich and Brown (1979) for the REML approac...
TL;DR: Tables of critical values for a new multivariate goodness-of-fit test introduced by Foutz are presented and details of the improved asymptotic approximation and evaluation of its accuracy are given.
Abstract: We present tables of critical values for a new multivariate goodness-of-fit test introduced by Foutz. Some details of our improved asymptotic approximation and evaluation of its accuracy are given. An example showing the application of the method is given.
TL;DR: In this article, a class of statistical tests for testing the hypothesis H 0Δ=0, where Δ is a location parameter, is proposed, defined in terms of the signed ranks of the sample data.
Abstract: A class of statistical tests for testing the hypothesis H 0Δ=0, where Δ is a location parameter, is proposed. The test statistic is defined in terms of the signed ranks of the sample data. Il is shown, by the method of moment matching, that under the null hypothesis test statistics in the proposed class have approximately a t-distribution. A Monte Carlo simulation study is presented inorder to compare the significance levels and power of these tests with numerous other tests.