Mplus and the Appropriate Analysis of Ordinal, Non-Normal, and Hierarchical Data in Social Work Research

This workshop will demonstrate how social work researchers can use Mplus software to simultaneously address three common data problems in social work research: ordinal level data, non-normal distributions, and cases that are nested in neighborhoods, schools, agencies, or other units. Rationale: A substantial proportion of the data used in social work practice and research is collected with ordinal level scales (e.g., sets of questions with Likert scale response options). Data collected with high quality scales are best analyzed in the latent variable framework, but default structural equation procedures assume interval level or higher measures and normal distributions. In addition, not all SEM programs have the ability to appropriately accommodate the nested nature of data. Ignoring common undesirable data characteristics in statistical analyses threatens the accuracy of specific findings and the credibility of social work research in general. Content: The workshop will teach participants appropriate analysis strategies for categorical and ordinal level scales, non-normal distributions, and nested data. These three data issues can be handled easily in the software program Mplus. The workshop will teach (a) the advantages of structural equation modeling (SEM) with latent variables when multiple items are used to measure social work constructs, (b) the appropriate analysis procedures for categorical, ordinal level, and/or non-normal items, and (c) the appropriate procedures for analyzing nested data. Basics of SEM will be covered as a foundation for the discussion of how Mplus handles ordinal, non-normal, and grouped data. Pedagogical techniques will include: 1. Introduction, 2. Interactive discussion of participants' current analysis strategies for ordinal variables and/or group data, 3. Presentation of the advantages of (a), (b), and (c) above, including comparisons of results obtained with appropriate and inappropriate analysis techniques, 4. Demonstration of analysis examples using correct procedures in Mplus, 5. Interactive discussion of participants' data and possible analysis strategies, and 6. A handout including the presentation slides, references, and sample Mplus syntax that can be modified for participant's analyses. Audience: The presentation will be especially beneficial to participants who have conducted exploratory factor analyses, confirmatory factor analyses, or general SEM with latent variables, but who have not addressed the measurement level, normality violations, and nested characteristics of their data. It will also be useful to anyone who collects data with ordinal scales, those who have some knowledge of regression or path analyses, and anyone developing measures for research or practice. PhD students planning structural equation or confirmatory factor analysis dissertations are welcome. Goals: At the end of the workshop, participants will understand how important it is to use appropriate analysis techniques for ordinal, non-normal, and grouped data. They will understand basics of structural equation modeling with latent variables. They will be prepared to conduct analyses with their own data using appropriate Mplus techniques; specifically, they will be equipped with sample Mplus syntax and knowledge of how to adapt it to their own analyses. Knowledge of the procedures covered in the workshop will improve the ability of social workers to publish their studies in the most rigorous journals.