Understanding Smoke Risk Perceptions in Idaho through AMOS-Based CFA and SEM

Author: Shahriar Md Arifur Rahman 

In this spring semester, I was continuing my previous work on Idahoans’ perception of risk from wildfire smoke. Building on the survey conducted last year, this phase of the research focused on analyzing how different psychological and environmental factors—such as dread, knowledge, controllability, and particulate matter (PM2.5) exposure—shape protective behaviors. Our primary aim was to validate a theoretical model of risk perception using the psychometric paradigm, especially examining how demographic differences and exposure history influence behavioral responses to wildfire smoke. 

A significant part of this semester was devoted to analyzing our dataset of 506 respondents using SPSS AMOS. The objective was to validate our theoretical model through Confirmatory Factor Analysis (CFA) and to explore pathways through Structural Equation Modeling (SEM). CFA allowed us to validate the underlying measurement model by examining how well observed variables (survey responses) represented latent constructs like dread, knowledge, and controllability. We checked the key validity criteria through the Fornell-Larcker criterion and meeting all the requirements helped us refine a measurement model that was both statistically sound and conceptually meaningful.  

Figure 1. Path analysis through SEM in the AMOS environment. 

Once validity and reliability were established, we tested multiple models to understand how different factors predict protective behaviors. Model 4 offered the most nuanced insights by splitting the sample between amenity migrants and long-term Idahoans. Among migrants, dread, knowledge, self-efficacy, and recent PM exposure all significantly predicted behavior. For long-term residents, only dread and self-efficacy were significant, confirming a degree of behavioral inertia and possibly a shifting baseline in risk perception. 

The structured visualization in SPSS AMOS helped communicate complex relationships between psychological, environmental, and behavioral variables clearly, supporting our broader aim of understanding how wildfire smoke risk is internalized and acted upon across demographic divides. Going forward, these models will inform the manuscript draft and serve as a foundation for broader discussions around air quality communication and adaptive behavior in the face of smoke related hazards.