Statistical analysis becomes to undertake an important role in the business systems of company, then such systems may build in statistical functions. In this case, it is possible to develop software of statistical functions. But I recommend to apply existing software such as SAS or S-PLUS etc. as considering efficiency of development and reliability of the software. In this paper some systems which incorporate existing statistical software are introduced, the advantages and disadvantages are discussed, and desirable new functions to statistical software are suggested.

View full abstract

In this paper, we consider analysis of categorical data that may plausibly contain over/under dispersion. We focus our attention to covariate effects estimation and evaluation. Several methods, such as maximum likelihood approach based on multinomial distribution and its parametric extension, i.e. Dirichlet-multinomial (DM) distribution and quasi-likelihood methods or generalized estimating equations with use of partial structure of the DM distribution, as well as computer intensive methods, such as Jackknife methods and Bootstrap methods are reviewed. Their relative advantages and limits are examined and discussed via their applications to some actual data sets.

View full abstract

Computing statistics approach to time series analysis is shown. Various important problems in time series analysis such as prediction, interpolation, decomposition and the parameter estimation can be unified as the state estimation problem of a properly defined state space model. For the standard linear Gaussian state space models, the state estimation can be achieved very efficiently by the famous Kalman filter and fixed interval smoothing algorithms. By intensive use of computers, a similar recursive computation for very general nonlinear non-Gaussian state space model can be developed. In this paper, numerical methods based on numerical integration and Monte Carlo approximation are shown. A self-organizing state space model for simultaneous estimation of the state parameters and the structural parameters is also shown. As numerical examples, estimation of changing volatilities of seismic data and stock price data are considered.

View full abstract

The spatial scan statistic is a method of detection and inference for the zones of significantly high or low rates based on the likelihood ratio. Kulldorff (1997) detected the hotspots based on spatial scan statistic with Binomial and Poisson models. The circular window zone for scanning is defined around one cellular (county) seat. The zone consists of counties whose county seat exists within the circle. Thus this approach has the properties to find the circular cluster as the candidate of hotspots. The echelon (Myers et al., 1997) is useful technique to study the topological structure of a surface in the systematic and objective manner. The echelon dendrogram represents the surface topology of cellular data and hierarchical structure of these data. The candidates of hotspots are given as the top echelon in the dendrogram. The purpose of this paper is to detect the any shapes of hotspots based on echelon analysis and spatial scan statistics. We demonstrate the procedure to find the hotspots based on newly proposed technique for sudden infant death syndrome (SIDS) data for each of the counties of North Carolina for the periods of 1974-1984. We also detect the hotspots of cells for r × c two-way ordered categorical data in contingency table with the example.

View full abstract

Through intending to research "statistical science", studying the main subjective involved and doing statistical data analysis in practice, we have felt that it is necessary to have something like a philosophy in doing such a thing. In statistical science, an ounce of practice is worth a pound of theory. In fact, statistical science is evaluated by its practical usefulness and the theory reaches a deadlock when it is completed. An own philosophy based on "wisdom or sophia by experiences"plays an active part such a situation. Originality cannot exist in which a practice does not play an active part. In fact, the teaching of "Learning by a fact makes originality" is to represent the essence of statistical science. Here we introduced some examples extracted by famous literatures in which the results and productive findings are more fruitfully obtained by our reanalysis, reexamination, and reinvestigation rather than the methods used in the papers. In the process of "quasi-data analysis" or "anatomy of some literature examples", we briefly introduced some methodologies developed in our limited experience. Furthermore, as some rules of thumb, we provided some instruction in the statistical data analysis process which consisits of four elements, namely, data, methodologies, analysts or users, and environments.

View full abstract

Recently, a multi-agents system has been discussed in the field of adaptive control. An autonomous clustering is considered to be a multiple agents system which constructs clusters by moving each pair of objects closer of farther according to their relative similarity to all of the objects. In this system, the objects correspond to autonomous agents, and the similarity relation is regarded as the environment. Defining a suitable action rule for the multi-agents, the clusters are constructed automatically. In this paper, we discuss the ability of the detection of clusters by the autonomous clustering technique through the concrete examples.

View full abstract

In this paper, I propose new teaching of statistical education that is composed of three steps. First step is that students calculate elementary statistics (mean, median, mode, range, range of quantile, standard deviation, skewness, kurtosis, percentile etc), correlation coefficients, cross tabulation and regression model of small data such as 4 cases by two variables in addition to draw histogram and scatter plots. Second step is that they learn how to analyze data by statistical software packages such as JMP, Statistica, SPSS, S and SAS. Third step is that they collect data from HP or CD-ROM and analyze it and write a repot by themselves. If they can take such new education as a start line about statistics, they can use their practical knowledge to their work after they graduated from university.

View full abstract

Before, softwares for statistical analysis were used only by professionals, but nowadays use of business people go on increasing. Instructions to the software are changing into end-user programming language and graphical user interface. Because of enpanding data scale, these softwares will be able to hand larger scale data, over 32bit cpu and memory limitation. By standardization of data format and data exchange protocol, we will be able to merge different kind of data from many data source easily, and to analysis integrated data sets.

View full abstract

The present paper reviews recent researches on influence diagnostics in multivariate methods, which deal with the evaluation of the influence of observations on the results of analysis or the detection of singly and/or jointly influential observations, and discusses two topics in details. The first topic is Cook's local influence in PCA and its relationship with our multiple-case diagnostics based on the influence function approach . Here the local influence is derived where the parameters of interest are contained in equality contraints but not in the likelihood function. So far the local influence has been derived in the cases where the parameters of interest are contained in the likelihood functions. But, as shown in PCA, the local influence can be derived not only in the cases where the parameters of interest are contained in the likelihood functions but also in the cases they are contained in equality contraints. The equivalence holds between the results of the local influence approach and the results of multiple-case diagnostics in the influence function approach. The second topic is how to standardize the influence functions with a continuous argument such as the influence functions of weight functions in the functional data analysis. We transform such continuous influence functions to discrete influence functions by sampling from the range of the continuous argument with appropriate intervals, and then apply Mahalanobis-type standardization. Our basic idea is to define the standardization in the continuous case as the limit when the intervals approach zero and the number of sampling tends to infinity. We can prove that when the number of sampled points is larger than a certain number, the statistics in the multiple-case as well as single-case diagnostics become constant and the values are the same with the results when we apply influence diagnostics to the coefficient vectors of the basis function expansions.

View full abstract

In this paper, we survey methods for curve fitting for multidimensional data without external criterion. Curve fitting is a fundamental techunique for data analysis. Principal Components Analysis find out a line or linear subspace fitted for the data. But, it is not easy to get nonlinear curve fitting. We introduce three methods for nonlinear curve fitting; Generalized Principal Components Analysis, Algebraic Curve Fitting, and Principal Curve.

View full abstract

The rapid development on computer technologies gives us a lot of variations when we use a computer in data analysis. In addition, the computer network like the Internet has become popular all over the world. This means most persons can collect any kind of data easily when they want to analyze the data properly. To utilize the powerful resources and analyze huge data, we survey recent variations on fast computing environments, review some studies related to parallel computing mainly, and offer some discussions in terms of computational statistics.

View full abstract

The roles and research area of Computational Statistics have changed by the diffusion of the Internet. The change spans widely, not only research field of computational statistics but also education methods, the statistical information, Internet survey, and statistical analysis system. In this paper, I will report the situation in recent years for activities of computational statistics, such as information of computational statistics, Internet survey, education for statistic, statistics analysis system. Furthermore, I will describe future activity of these areas.

View full abstract

Positive-definite symmetric matrix is useful for various multivariate analyses. The latent roots and vectors of the symmetric matrix are also important statistics. Covariance matrix is typical for positive-definite symmetric matrix. In this paper we deal with the hypothesis testings for the latent roots and vectors of covariance matrix without normality. Three hypotheses are considered for the case of one population and two populations, respectively. We explain that Wald criteria are effective for these problems and the criteria are derived by using asymptotic covariance matrix for latent roots and vectors.

View full abstract

Hierarchical clustering algorithms are generally based upon a (dis)similarity that is assumed to be symmetric between object pairs. However, the (dis) similarity used in actual analysis is asymmetric. Therefore, to analyze the asymmetric (dis) similarity data the researcher must perform a somehow symmetrization of his original proximity values in the beginning. On the other hand, the idea that the asymmetry has elemental meaning and the researcher must analyze the data given by using algorithm depending on the asymmetry was suggested Hubert (1973) proposed "min and max clustering" for the asymmetric similarity. He symmetrized the data matrix in the beginning and analyze using the "min and max clustering algorithm". Fujiwara (1980) extend the Hubert's (1973) algorithm. He suggested the researcher should not perform symmetrization and should analyze the original asymmetric data matrix. Algorithms proposed in these two papers were extended the single linkage algorithm and the complete linkage algorithm to the asymmetric clustering algorithm. Okada and Iwamoto (1996) proposed the weighted average algorithm for asymmetric (dis) similarity. In those papers, they defined algorithms by deciding two steps, (i) selects the objects to be combined and (ii) updates the (dis) similarity between the objects, and not proposed uniformly. In this paper, we define an extended updating formula to handle a profusion of asymmetric hierarchical clustering algorithms uniformly in the same manner as the symmetric one by Lance and Williams (1967). Extended dendrogram for representation of the result of analysis for asymmetric data is also proposed.

View full abstract