Structural equation modeling and the analysis of long-term monitoring data

Abstract

The analysis of long-term monitoring data is increasingly important; not only for the discovery and documentation of changes in environmental systems, but also as an enterprise whose fruits validate the allocation of effort and scarce funds to monitoring. In simple terms, we may distinguish between the detection of change in some ecosystem attribute versus the investigation of causes and consequences associated with that change. The statistical framework known as structural equation modeling (SEM) can contribute to both detection of changes and the search for causes. This chapter summarizes some of the capabilities of SEM and shows a few ways it can be used to model temporal change. Because of its ability to test hypotheses about whether rates of change are zero or nonzero, it can be used for change detection with repeated-measures data. As more of the capabilities of SEM are presented, its capacity for evaluating causal networks is highlighted. Here is where its potential for making a unique contribution to the analysis of long-term monitoring data is revealed. Thus, one?s primary motivation for using SEM with monitoring data will be to investigate hypotheses about what factors may be driving change. In this chapter, it will be necessary to first introduce notation to describe the elements of structural equation models (SE models) so as to permit an unambiguous presentation of their various forms. The first part of the chapter works through the fundamental features of models of increasing complexity, while the second part of the chapter illustrates several of these possibilities using a real example.