In this paper, we report our attempt to construct a software system for time series model fitting, TMODEL, on a 16-bit personal-computor(PC-9800). This system has the following features : i) easiness to approach, ii) graphical ability and iii) extendability and modifiability of programmes. To construct our software system for time series data analysis, we suppose a user having some fundamental knowledge about time series analysis, so that the user can easily modify and/or extend programmes by himself. As an illustrative example for time series model fitting by our system, we made a short-term prediction of the famous Canadian lynx data. In the process of our study of constructing the software, system, it occurred to us that the system itself should have some knowledge about statistical time series analysis which can help a user to read a graphical features of time series data and to choose a next proper action for him to take.
Key words: Time series model fitting, Prediction, ARMA-model, Pretreatments of time series dataIn an imbalanced one-way ANOVA model with homogeneous variance, we examine the performances of Tukey's multiple comparisons procedure. The efficiency of Tukey-Kramer's approximate method to the exact method is evaluated based on some numerical results, which are focused on both of the critical value and the p-value. When the sample size is severely imbalanced, we recommend to use the exact calculation since Tukey-Kramer's method is too conservative in such a case. We conclude that Tukey-Kramer's method is a good approximation for moderately imbalanced cases that are frequently encountered in many practical applications.
Key words: Critical value, Multivariate t-distribution, Pairwise multiple comparisons, P-value, Studentized range
The problem of deriving latent ordered classes is discussed in the latent structure analysis, where these classes are assumed to be ordered based on binary items which are regarded as indexes for one latent trait.
Up to now, the problem of ordered classes has been discussed only as the result of interpretation of the derived latent classes in the general latent class analysis. Thus, the ordinary latent class model does not have any concepts of latent ordered classes.
In this paper, the assumption of latent ordered classes or unidimensional concepts is incorporated into the model itself. Unlike the ordinary latent class model, the aim of this paper is to analyze data by the aid of a structured model. As the structured model, a logistic model is adopted and the algorithm of ML estimation for this model is constructed based on EM algorithm.
The aim of this paper is to get a multidimensional scale for asymmetric dissimilarities. As a first step of this scaling, a distance representation of asymmetric dissimilarities is performed using an asymmetrical metric function(asymmetric distance function). For this purpose, Minkowski space is discussed as a model of the space which has the asymmetrical metric function. In this space, a concrete class of the asymmetrical metric function is obtained using a concept of oval or ovaloid. We applied these metric function to MDA(Minimum Dimension Analysis)for a sociometric data.
Key words: Multidimensional scaling, Distance representation, Asymmetric dissimilarities, Minkowski metric function, OvaloidIncomplete U-statistics are obtained by sampling summands of a U-statistic. In two-sample cases, a problem of how to derive an optimal incomplete U-statistic is more complicated than that in one-sample cases. In this paper, we consider a sampling scheme to derive incomplete U-statistics in two-sample cases and present an optimal one in four cases which are useful in the practical situations. It's found that the optimal procedures ease us of heavy computational burden with little loss of information.
Key words: Incomplete U-statistic, Efficiency, Optimality, Computation of U-statisticWe review smoothing-intensive methods of regression, which include PPR, GAM, ACE and so on. We reanalyze three sets of data cited from published literatures by applying GAM and then illustrate practical behaviors of these methods based on smoothing. These methods can give attractive findings which may not be obtained by usual analyses but provide difficulties in their implementation and interpretation.
Key word: Smoothing, Computer-intensive method, PPR, GAM, ACE, GLM, LOWESS, Regression, Discrimination