When it comes to analyzing data from a 2×2 factorial vignette study, it’s crucial to choose the right statistical approach. In this post, we’ll dive into the challenges of analyzing ordinal dependent variables (DVs) with violated parametric assumptions and explore strategies for overcoming these obstacles.
A recent study investigated public stigma towards comorbid health conditions, specifically epilepsy and depression. The study design involved a 2×2 between-subjects factorial vignette survey with 225 participants, where participants were randomly assigned to one of four vignettes: Control, Epilepsy-Only, Depression-Only, and Comorbid. The dependent variables were measured using two ordinal scales: the Attribution Questionnaire (AQ) and the Social Distance Scale (SDS).
The main goal of the study was to determine the presence and nature of stigma towards the comorbid condition. However, the analysis was complicated by the violation of parametric assumptions, specifically homogeneity of variance and normality of residuals.
So, what’s the best approach for analyzing these data? One option is to use ANCOVA, but this requires treating the total scores as continuous variables, which may not be suitable for ordinal data. Another option is to use ordinal logistic regression, which can handle the ordinal nature of the data but may require multiple comparisons and adjustments for family-wise error rates.
To visualize the nature of stigma, stacked bar charts can be used to show the proportion of responses for each Likert category across the four conditions. However, it’s essential to determine the best way to present the results, whether it’s through proportional odds ratios, chi-square tests, or ANOVA.
Ultimately, the choice of statistical approach depends on the research question, study design, and data characteristics. By understanding the strengths and limitations of different methods, researchers can make informed decisions about how to analyze their data and present their findings in a clear and meaningful way.