Showing papers in "Statistica in 1991"
TL;DR: In this paper, the influence function of the Gini concentration ratio is determined and the resulting function is used to verify the robustness properties of the functional, which proves sensitive to the presence of outliers.
Abstract: In this paper the influence function of the Gini concentration ratio is determined. The resulting function is used to verify the robustness properties of the functional, which proves sensitive to the presence of outliers. The local shift sensitivity is then determined and it is shown to be bounded under the assumption that income is a continuous random variable. Finally the influence function is used in the approximation of the bias of Gini's concentration ratio due to grouped data.
57 citations
TL;DR: Though the proposed estimator is biased, the bias can be reduced by choosing the parameters suitably, and the situations, where the suggested technique fares well as compared to original unrelated question randomized response procedure have been identified.
Abstract: In this study, an optional randomized response strategy has been proposed using non-stigmatized attribute. The proportion of the non-sensitive attribute has been assumed to be known. Though the proposed estimator is biased, the bias can be reduced by choosing the parameters suitably. The situations, where the suggested technique fares well as compared to original unrelated question randomized response procedure have been identified.
14 citations
TL;DR: In this paper, an alternative randomized response sampling procedure for estimating the proportion of human population possessing a sensitive characteristic is considered, which is unbiased and takes inadmissible values outside the range [0, 1] less frequently as compared to the Warner's (1965) usual estimator.
Abstract: In the present study, an alternative randomized response sampling procedure for estimating the proportion of human population possessing a sensitive characteristic is considered. The proposed estimator is unbiased and takes inadmissible values outside the range [0,1] less frequently as compared to the Warner's (1965) usual estimator.
12 citations
TL;DR: In this article, the authors investigate the robustness of significance tests to error-autocorrelation in the linear regression model, and derive the limiting significance levels as autocorecorrelation increases.
Abstract: We investigate the robustness to error-autocorrelation of significance tests in the linear regression model, and derive the limiting significance levels as autocorrelation increases. The main result is that the degree of nonrobustness of the tests depends crucially on the regressors of the model.
6 citations
TL;DR: In this article, a recursive definition of regularity is proposed, able to unify the three Elementary Particle Statistics, which can be considered as an unified approach to predictive inductive inferences from the particular point of view of Statistical Physics.
Abstract: A unified derivation of Elementary Particle Statistics is hindered by Pauli's Exclusion Principle, that limits the possible states of the physical system in a peculiar way. It follows that if the Principle does not hold (Maxwell-Boltzmann e di Bose-Einstein cases) all occupation vectors are regular, so long as they are allowed by the macroscopic parameters. On the contrary, if the Principe holds (Fermi-Dirac case), the regularity domain of the probabilistic description is narrowed by the Principle itself. In this paper a recursive definition of regularity is proposed, able to unify the three Statistics. In this general formulation Pauli's Exclusion Principle can be found as a particular case. The paper is based on the notion of stochastic process with exchangeable increments. It can be considered as an unified approach to predictive inductive inferences from the particular point of view of Statistical Physics.
5 citations
TL;DR: In this article, an accelerated sequential procedure is proposed to save the number of sampling operations we would have observed had the purely sequential procedures been used, and the method is outlined as it pertains to the point estimation of the unknown normal mean.
Abstract: An accelerated sequential procedure is proposed to save the number of sampling operations we would have observed had the purely sequential procedures been used. The method is outlined as it pertains to the point estimation of the unknown normal mean. We supplement our theoretical findings with simulations to study the moderate sample size performance of the proposed procedure.
5 citations
TL;DR: In this article, a general procedure of parametric estimation for distributions having characteristic function with polar coordinates in closed form is proposed, which generalizes the method of Koutrouvelis (1982a) for estimating the parameters in the special case of Cauchy distributions.
Abstract: A general procedure of parametric estimation is proposed for distributions having characteristic function with polar coordinates in closed form. The procedure generalizes the method of Koutrouvelis (1982a) for estimating the parameters in the special case of Cauchy distributions. The resulting estimators are asymptotically equivalent to the estimators proposed by Feuerverger and McDunnough (1981b) and serve as a basis for chi-squared goodness-of-fit tests. For symmetric families the employment of polar coordinates of the empirical characteristic function is shown to have definite advantages in both estimation and testing.
3 citations
TL;DR: In this article, the author examines techniques already known in literature, that can be derived from FAMA, adding new tools of analysis, and traduces from FActorial Matrices Analysis new methodologies to overcome problems connected with some of the known techniques.
Abstract: In this paper after introducing the FActorial Matrices Analysis (FAMA ) in order to study a three-way matrix (Vichi, 1990), the Author examines techniques already known in literature, that can be derived from FAMA, adding new tools of analysis. Furthermore the Author traduces from FAMA new methodologies to overcome problems connected with some of the known techniques.
2 citations
TL;DR: In this article, it has been shown that one of the two available definitions of LP is hardly acceptable outside a strictly Bayesian context, and that the LP is implied by other more fundamental principles which most statisticians tend to accept.
Abstract: Two points have been raised about the Likelihood Principle (LP). The first relates to the definition of LP: it has been shown that one of the two available definitions is hardly acceptable outside a strictly Bayesian context. The cue for the second issue is taken from Birnbaum's proof (and other similar proofs) that the LP - in its strong version - is implied by other more fundamental principles, which most statisticians tend to accept; some of these principles or axioms, after subjection to severe critical discussion, have been recognized as inadequate for general agreement. Moreover, it is shown that Birnbaum's proof - as well as others with the same demonstration structure - is logically false.
2 citations
TL;DR: In this article, Lagrangian or "random flight" models of dispersion are used to compute the trajectories of a large number of marked fluid particles and the ensemble average of such trajectories allows an evaluation of the concentration of the marker.
Abstract: Enviromental control problems demand an increasing care in modelling diffusion in real turbulent flows In this paper we will deal with a problem arising with the so-called "Lagrangian" or "random flight" models of dispersion In such models some statistical informations about flow turbulence (mean value, variance and possibly skewness of the velocity component distributions) are used to compute the trajectories of a large number of marked fluid particles The ensemble average of such trajectories allows an evaluation of the concentration of the marker These dispersion models are specially appropriate for inhomogeneous turbulence fields, such as it occurs in the atmosphere over hills or near large buildings
2 citations
TL;DR: In this article, the maximum likelihood estimation of parameters of a Power Series Distribution (PSD) when the sample is type I progressively censored with different parameters at different stages of censoring is investigated.
Abstract: The paper deals with maximum likelihood estimation of parameters of a Power Series Distribution (PSD ) when the sample is type I progressively censored with different parameters at different stages of censoring. The discrete probability model characterized by PSD under the scheme of type I progressive censoring is motivated with reference to studies in life testing and reliability and the likelihood estimates of Geometric distribution as obtained by Patel and Gajjar (1990) for two-stage progressive censoring are exemplified as special cases.
TL;DR: In this article, the use of M-estimators for a sample containing censored observations is discussed, and a simple and intuitively acceptable extension of a generic M estimator is considered.
Abstract: In the first part of the paper, the use of M-estimators for a sample containing censored observations is discussed. In particular, a simple and intuitively acceptable extension of a generic M-estimates is considered. The second part discusses robust estimation of the parameter of an exponential model. This part contains directly applicable results and illustrates the construction of the estimators previously suggested.
TL;DR: The author proposes to determine some general relationships linking the functions that express the effectiveness of a sampling plan in lot acceptance sampling by introducing a function which provides the expected number of units, to be drawn from a lot of given quality, with reference to accepted lots only.
Abstract: In this paper the author proposes to determine some general relationships linking the functions that express the effectiveness of a sampling plan in lot acceptance sampling. The result is obtained through the introduction of a function which provides the expected number of units, to be drawn from a lot of given quality, with reference to accepted lots only. This function, together with the OC function, appears in the expression of the ASN, ATI and AOQ functions, which provide the economic design of a sampling plan. In addition, the author defines two functions by introducing some costs per unit of the inspection, substitution and utilization of defective items. The first one expresses the producer's expected gain from the application of a sampling plan to replace one hundred per cent inspection of lots, the second one expresses the consumer's expected gain from the application of a sampling plan to replace the acceptance of lots without inspection. For whichever type of sampling plan, single, double, multiple or sequential, it is sufficient to determine the form of the first function mentioned in the opening paragraph and the OC function, to express the remaining functions. Some examples are provided to clarify the proposed procedure. Of particular importance is the application in the case of sequential sampling plans where, through the use of some results related to random walks with two absorbing barriers, it is possible to obtain approximate expressions of all the above functions.
TL;DR: In this article, the authors extended Harter and Moore's results to the estimation of the parameters of type-I asymptotic distribution of m -th largest observation of several samples.
Abstract: Let X be a random variable following the first asymptotic distribution function of extreme values with location parameter u and scale parameter b , b > 0. Harter and Moore (1968) presented the MLEs of u and b from censored samples with lowest r 1 and largest r 2 sample values being censored in a sample of size n . In the present study these results are extended to the estimation of the parameters of type-I asymptotic distribution of m -th largest observation of several samples. The method of maximum likelihood with a sample of size n having r 1 smallest and r 2 largest sample values censored, is used for estimation. The asymptotic variance-covariance matrix of the MLEs is obtained by inverting the information matrix, whose elements are the negatives of the limits, as n → ∞, of the expected values of the second partial derivatives of the likelihood function.
TL;DR: In this article, an autoregressive continuous time process is used to predict an observed random function (Z (t), t ~ R) on a time-interval, and the predictor is constructed as a function of an estimator of the correlation function of Z.
Abstract: In order to predict an observed random function (Z (t), t ~ R) on a time-interval, we modelise Z by an autoregressive continuous time process, and construct the predictor as a function of an estimator of the correlation function of Z. Simulations and applications to experimental data sets are given.
TL;DR: In this article, the small-sample properties of estimator in linear models using the exponential transformation are investigated via a Monte Carlo experiment, and at different values of the transformation parameter they have found considerably different qualitatively results.
Abstract: The small-sample properties of estimator in linear models using the exponential transformation are investigated via a Monte Carlo experiment. At different values of the transformation parameter we have found considerably different qualitatively results.
TL;DR: The authors argued that Bayesian theory, which allows judgement samples, provides an intuitively correct basis for investigating population properties, and concluded that the principle of random sampling is not well founded.
Abstract: Judgement sampling in market research and opinion polling is standardly criticized as unsatisfactory on the grounds of its subjectivity, and it is widely held that random sampling is the only scientific means of investigating properties of large populations. This paper investigates these claims. It concludes that the principle of random sampling is not well founded. It also argues that Bayesian theory, which allows judgement samples, provides an intuitively correct basis for investigating population properties.
TL;DR: The on-line rupture detection and classification problem is solved by means of a multiple decision procedure which enables the user to localize the rupture area.
Abstract: In dynamically control field, the effectiveness of closed loop control is based on good knowledge of system parameters; it is then important to detect parameters changes, chiefly in off-line estimation field. According to the physical structure of plants represented by closed loop autoregressive models the system parameters may be divided into four classes: level and dynamic of transform system, level and dynamic of control system. This gives rise to four kinds of parameter ruptures, where rupture means step-change from one state to another. In the present paper the on-line rupture detection and classification problem is solved by means of a multiple decision procedure which enables the user to localize the rupture area.