ABSTRACT
Structural Equation Modeling (SEM) is gaining attention in social work as an analytical method for studying social policy and testing theories for practice. This paper demonstrates a special SEM application to test a theory for research use in the human services. Data from 294 decision-makers in human service organizational units of federal, state, and local governments are used to show the appropriateness of using composites instead of multiple indicators in a SEM model with latent variables. The step-by-step description of a two-stage modeling process is presented. Using a model generation analysis strategy, the results show that a nonrecursive model positing a reciprocal relationship between research use and diffusion could not be confirmed. Instead, a recursive model is uncovered with research use leading to research diffusion. In addition, perceived quality of research, positive interaction between researchers and decision-makers, and transfer intensity of disseminating research results through multiple media, produced direct and indirect effects on research use and diffusion. The limitations and strengths of this study using the special SEM application are discussed.