| This research describes the development of a microcomputer-based analytical system for the modeling of Markov chains. The system provides for the analysis of first-order stationary Markov chains that are regular, absorbing or periodic. The system allows the user to classify the chain, to test hypotheses concerning order (Markovity) and stationarity, and to calculate classical descriptive statistics for the stochastic process. Options are available for the determination of population heterogeneity as a function of discrete covariates using log-linear and/or linear categorical methods, thus providing a method of testing the equality of transition probability matrices. Other system components enable the user to import and export ASCII (American Standards Committee on Information Interchange) format data files (both with and without covariates), to list the data and cross-tabulate variables, and to conduct process simulation of perturbations to the model. The system is menu driven with prompts for user input. Output of all procedures can be directed to the screen, printer, and/or a disk file (in standard ASCII format).; The system was programmed using the GAUSS matrix language (version 2.0) and was designed to run on IBM-PC compatible microcomputers with hard disk drives and 80286/80287 or 80386/80387 microprocessors. The system is available as a p-code compiled program with a GAUSS run-time module. |
