Speaker: Arlie Beliveau, York University
Department of Psychology

Title: Reevaluating the Essen CES: A case study from the Waypoint Centre for Mental Healthcare

Abstract: In the past, researchers have used Confirmatory Factor Analysis (CFA) with Pearson correlations to model the ESSEN CES -- a therapeutic climate survey with fifteen 5-point Likert-scale test items that load onto three latent variables: Patient Cohesion (PC), Experienced Safety (ES), and Therapeutic Hold (TH). This approach may be problematic because the test items appear categorical in nature and may violate the statistical assumptions of the analysis. Using any Maximum Likelihood (ML) based test, such as factor analysis, to model categorical data can produce attenuated or diluted estimates of the correlations among indicators, and incorrect test statistics, standard errors, and parameter estimates.

I screened data from the 2014 Essen CES collected at the Waypoint Centre for Mental Health to see if it met the assumptions for CFA. There are problems including possible categorical items, non-normal data (skewness and kurtosis), and missing data. If it makes theoretical sense to consider the test items as having continuous underlying traits, then it would be appropriate to utilize a more robust polychoric analysis. If the items should be considered truly categorical, then perhaps Item Response Theory would be better suited. This paper is a preliminary report of my analysis and is part of my Minor Area Paper.