Submitted to: Proceedings Of Animal Stress Workshop
Publication Type: Proceedings
Publication Acceptance Date: 12/21/1995
Publication Date: N/A
Citation: N/A Interpretive Summary: This paper was prepared for presentation to animal scientists, engineers, and animal physiologists with a common interest in animal stress. Because of the broad interests of the audience, the subject was developed using several foundational concepts of modeling and interpreting dynamic systems. To do this, three main topics were addressed: 1) Introduction of time series measurements. Proper sampling procedures and precautions are necessary to accurately measure dynamic systems. Methods of minimizing sample measurements through filtering and of determining minimum sampling rates were presented. 2) Discussion of the interpretation of models. Several qualitative measures of time series measures were given that are useful for describing dynamic systems. An introduction to computer modeling and other mathematical analysis techniques provided direction for guiding linear modeling projects. 3) Alternative nonlinear and multi-variable models were introduced. Alternative modeling techniques discussed were adaptive models, neural networks, and fuzzy systems. Such techniques are useful for situations where simple linear models are inadequate to track more complex dynamic systems. The discussion in this paper provides a jumping off point for persons interested in pursuing more advanced models applicable to evaluating animal stress measures.
Technical Abstract: This review article introduces concepts and methods of measuring and modeling dynamic systems. The first topic covered is time series measures. The need for proper sampling procedures to prevent aliasing is described. Power spectrum analysis was also introduced as a means of determining appropriate sampling rates. The second topic covered was interpretation of linear models. Qualitative descriptors of linear model dynamic responses such as lag time, rise time, overshoot and settling time were presented. The use of the Laplace and Z transforms to select appropriate linear models and to numerically simulate real systems were presented. Finally, a brief description was given of non-linear and/or adaptive models, including adaptive model estimators (such as a Kallman filter), neural networks, and fuzzy system models.