Evaluating the effects of toxicants in ecosystems is difficult despite numerous attempts to develop field and laboratory tests. The problem appears to be the lack of an analytical methodology capable of taking advantage of the available experimental designs. Therefore, we propose a technique for modeling ecosystems-linear structural modeling with LISREL. LISREL is a path analytic technique that is more flexible than classical path analysis. Modeling with LISREL involves placing ecosystem structure and function into a framework of concepts and indicator variables. Concepts are theoretical constructs that are placed into a cause-and-effect network to reflect true ecosystem structure. Concepts are @indicated" by indicator variables; these are the measured variables in the ecosystem. LISREL incorporates measurement error into the modeling process by establishing a portion of the variance of each indicator variable as measurement error. The framework of concepts and indicators in a cause-and-effect network becomes a hypothesis that is tested using LISREL. LISREL also provides a determination of the total, direct, and indirect effects of the variables on each other. A measure of ecosystem stability is provided as part of the modeling process.
© 1991 by the Ecological Society of America.