Category: Addictive Behaviors
Keywords: Addictive Behaviors | Research Methods | Dissemination
Presentation Type: Symposium
A primary goal of statistical modeling is to convey a cogent and compelling story about relationships observed in data. In structural equation modeling research, this is most commonly done through tables, path diagrams, and in-text reporting of statistical results. These methods of reporting are typically geared toward researchers with quantitative expertise and are commonly difficult for other research stakeholders and consumers (e.g., clinicians, patients, policymakers) to interpret, an issue of critical importance for substance use data. In contrast, effective use of data visualization may help treatment researchers more effectively convey research findings to broader audiences of stakeholders. This presentation aims to illustrate how researchers who use SEM and other latent variable modeling approaches may expand their use of data visualization to illustrate their findings to a broader audience of stakeholders. The presentation will first review the overarching aims of visual communication of data, and will discuss how existing research on visual communication, information processing, and human-centered design can inform decisions on the aims and designs of data visualizations. The presentation will illustrate how common structural equation models can be conveyed graphically through different formats and layouts, including multivariate scatterplots to visualize multivariate relations, radar graphs and dot-and-line graphs to illustrate factor loadings, and simple gauges to illustrate individual person-level outcomes. Methods for extracting, simulating, and presenting raw data, sample-average data, and model-implied data using the natural metric of study variables will also be presented. The presentation will highlight how the use of data visualizations can increase researchers’ ability to convey their intended messages to broader audiences. Illustrative examples will be included throughout the talk that draw on existing substance use treatment data. Key syntax from common statistical software and visualization tools that can support the generation of data visualizations will also be referenced.
Behavioral Research in Technology and Engineering Center, University of Washington
Friday, November 17
10:15 AM – 11:45 AM
The asset you are trying to access is locked. Please enter your access key to unlock.