The Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) is considered the “gold standard” for measuring symptoms for Obsessive-Compulsive Disorder (OCD) due to the high reliability and validity of the measure. Originally, the Y-BOCS was divided into Obsessive and Compulsive factors; however, literature on the factor structure of the Y-BOCS is inconsistent. Models range from one global factor to different interpretations of bi-factor models to three-factor models. Inconsistencies between models may be attributed to sampling error, including participants with subclinical OCD in some samples, and measurement error. In addition, many researchers treat the Y-BOCS measurement as an interval or ratio scale when it likely reflects ordinal measurement. Our paper has two primary aims. First, we compare the fit of the models proposed in the literature using a large sample from multiple sites of patients diagnosed with OCD. We also evaluate how the models can be improved and whether those improvements show evidence for convergent validity. We treat the Y-BOCS observations as ordinal data. Second, we evaluate measurement and structural invariance between genders. Additionally, we examine convergent validity of the factor structure of the best fitting model with subscales of the OCI-R. Data from five separate samples were combined into one dataset with 288 total participants all formally diagnosed with OCD. We selected several Y-BOCS factor models from the literature and used confirmatory factor analysis to evaluate goodness of fit indices on our pooled sample. Only one model approached acceptable goodness of fit indices. We considered the factors in this model and proposed a new factor model with a global factor (OCD) and two sub-factors (Obsessions and Resistance to Symptoms). Our model exhibited the highest goodness of fit indices which we further improved with modifications to our factor model. On invariance analyses, our model exhibited measurement invariance between genders and partial structural invariance. Additionally, the latent factors of our model exhibited convergent validity with all of the OCI-R subscales (except Ordering). Our model exhibited stronger goodness of fit indices with our data than existing models in the Y-BOCS literature and measurement invariance and partial structural invariance between genders. We recommend that future studies replicate the efficacy of our factor model using the Y-BOCS as an ordinal measurement.



College and Department

Family, Home, and Social Sciences; Psychology



Date Submitted


Document Type





invariance, factor structure, Obsessive-Compulsive Disorder