Abstract
Context: When working with health-related questionnaires, outlier detection is important. However, traditional methods of outlier detection (e.g., boxplots) can miss participants with “atypical” responses to the questions that otherwise have similar total (subscale) scores. In addition to detecting outliers, it can be of clinical importance to determine the reason for the outlier status or “atypical” response.Objective: The aim of the current study was to illustrate how to derive person fit statistics for outlier detection through a statistical method examining person fit with a health-based questionnaire.Design and Participants: Patients treated for Cushing's syndrome (n = 394) were recruited from the Cushing's Support and Research Foundation's (CSRF) listserv and Facebook page.Main Outcome Measure: Patients were directed to an online survey containing the CushingQoL (English version). A two-dimensional graded response model was estimated, and person fit statistics were generated using the Zh statistic.Results: Conventional outlier detections methods revealed no outliers reflecting extreme scores on the subscales of the CushingQoL. However, person fit statistics identified 18 patients with “atypical” response patterns, which would have been otherwise missed (Zh > |±2.00|).Conclusion: While the conventional methods of outlier detection indicated no outliers, person fit statistics identified several patients with “atypical” response patterns who otherwise appeared average. Person fit statistics allow researchers to delve further into the underlying problems experienced by these “atypical” patients treated for Cushing's syndrome. Annotated code is provided to aid other researchers in using this method.
Highlights
Detection of outliers on questionnaires is important in any field of research
It is the exact pattern of responses across items that will provide researchers with much richer and precise results for each individual
Patients in the present study were recruited by Tiemensma et al (2016), who examined different scoring options of the CushingQoL
Summary
Detection of outliers on questionnaires is important in any field of research (e.g., psychology, behavioral medicine). Outlier detection techniques, such as box-plots, determine outliers from total scores or subscale scores. While these methods for detecting outliers are the most common approach implemented for identifying “atypical” scores, there are certain forms of outliers that cannot be detected using this method. It is important to properly identify these participants, as their patterns may represent substantively interesting differences In this case, the total scores are not informative since they are comparable across participants. The total scores are not informative since they are comparable across participants Instead, it is the exact pattern of responses across items that will provide researchers with much richer and precise results for each individual
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
