Ordinary time-series analysis operates on time series data measured on interval scale. In the study of psychology, social science and management, we often encounter time-series data measured on nominal scale. In such cases, investigating the trend of response or selection is an important research activity. One of the ways to clarify the trend is visual expression of the trend. In this study, “time-series analysis of qualitative data”is developed to achieve this purpose. The basic idea behind the method is to assign numerical values or scores to categories so as to maximize correlation between the category scores and the function of time variable. This method includes five models: irreversible model, reversible model, circulative model, increase model, and decrease model. The method proposed in this study can be widely used as a tool for analyzing not only time-series data but also a set of data in which there are some external variables and a category data. An illustrative application to real data is presented to examine the availability of the method. Some technical problems and relation to other methods are also discussed.
The disaggregated logit model using random utility theory is most useful for the traffic demand estimation problem. Recent conventional studies of the disaggregated logit model have tried to improve estimation precision by reforming the utility function. Our research followed this method. In particular we considered improvement of estimation precision by combining the uncertainty of the travel time and disaggregated logit model. We applied the original model that reflects the personal attribute of subjective travel time. To put it concretely, we made the assumption that traveler's subjective distribution was affected by experience and made a model that estimated the parameters of subjective distribution(ex. standard deviation). Then, we did computational questionnaire survey in order to collect data for this model that analyzed the formation process of traveler's subjective distribution of travel time. Further we applied these survey data to a utility function of route choice problem, and concluded the model was fitting.
We are now observing quick and vast development of new non-linear methods for data analysis such as chaos and neural network in various scientific research fields. Those methods are also regarded as basic tools for the analysis in a still vaguely-defined research area called“complexity”It can be said that the development of such non-linear methodologies is no doubt inspired by recent rapid progress of high-speed computing facilities and graphical representation devices. The aim of the present article is two-fold. The first is to give a review of some recent development of chaos and artificial neural network(ANN)from a statistical viewpoint. Fundamental ideas and main features of chaos and ANN are shown with some illustrations of computer simulation. The second thing is, on the basis of past development of such methodologies, to mention some research topics which would be carried out near future in behaviormetrics and also in statistics, although such collection is of the author's personal view. In bibliography given at the end of the present article, research articles and review papers concerning non-linear methods appeared in main statistical journals after 1990 were collected and shown with short comments. Publications upon“complexity”for academic as well as for general readers are also listed. The list would be helpful for the reader to understand the basic idea of complexity and also to proceed his/her own research as well.