A Mokken scale analysis of the peer physical examination questionnaire

Abstract

BackgroundPeer physical examination (PPE) is a teaching and learning strategy utilised in most health profession education programs. Perceptions of participating in PPE have been described in the literature, focusing on areas of the body students are willing, or unwilling, to examine. A small number of questionnaires exist to evaluate these perceptions, however none have described the measurement properties that may allow them to be used longitudinally. The present study undertook a Mokken scale analysis of the Peer Physical Examination Questionnaire (PPEQ) to evaluate its dimensionality and structure when used with Australian osteopathy students.MethodsStudents enrolled in Year 1 of the osteopathy programs at Victoria University (Melbourne, Australia) and Southern Cross University (Lismore, Australia) were invited to complete the PPEQ prior to their first practical skills examination class. R, an open-source statistics program, was used to generate the descriptive statistics and perform a Mokken scale analysis. Mokken scale analysis is a non-parametric item response theory approach that is used to cluster items measuring a latent construct.ResultsInitial analysis suggested the PPEQ did not form a single scale. Further analysis identified three subscales: ‘comfort’, ‘concern’, and ‘professionalism and education’. The properties of each subscale suggested they were unidimensional with variable internal structures. The ‘comfort’ subscale was the strongest of the three identified. All subscales demonstrated acceptable reliability estimation statistics (McDonald’s omega > 0.75) supporting the calculation of a sum score for each subscale.ConclusionThe subscales identified are consistent with the literature. The ‘comfort’ subscale may be useful to longitudinally evaluate student perceptions of PPE. Further research is required to evaluate changes with PPE and the utility of the questionnaire with other health profession education programs.

Building Populist Discourse: An Analysis of Populist Communication in American Presidential Elections, 1896–2016

“…While Guttman/Mokken scales are most commonly used in the analysis of questionnaire responses, the scales can theoretically be used to analyze any concept in which the constituent "items" are intended to measure the same construct, but where each item is not considered equally "difficult" to achieve (van Alphen et al, 1994). MSA is based on a set of assumptions that make the resulting data theoretically interpretable (Vaughan and Grace, 2018). In particular, the data used to construct Mokken scales are assumed to be monotonic (the function is entirely non-decreasing) and unidimensional-if these assumptions are met, then the resulting scale can be interpreted as a measure of the overall construct that I am theoretically intending to measure.…”

Identifying Problematic Item Characteristics With Small Samples Using Mokken Scale Analysis

“…I used the GPCM to generate item responses because it allowed me to manipulate item-level discrimination that would result in violations of monotonicity and poor scalability. I generated data using the characteristics in Table 1; the design characteristics reflect recent real data analyses with MSA methods (e.g., Chou et al, 2017;Tillema et al, 2021;Vaughan & Grace, 2018). I simulated 500 replications of each condition.…”

“…I simulated five randomly selected items (hereafter, ''problem items'') in each condition to exhibit measurement problems; this value reflects the proportion of items that other researchers have identified as exhibiting problematic characteristics with MSA (e.g., Lee et al, 2017;Vaughan & Grace, 2018). I simulated problematic item characteristics within either a subset of the examinee location range or the entire range of examinee locations.…”

Comparing the Automatic Item Selection Procedure and Exploratory Factor Analysis in Determining Factor Structure

“…The higher the c coefficient used in item selection in the AISP is, the more strictly the process is carried out and the fewer item clusters are determined. The variable values of c in the AISP studies were compared with the selected item clusters (Chou, Lee, Liu, & Hung, 2017;Emons et al, 2012;Vaughan & Grace, 2018). A higher c criterion lower bound means that it is a powerful scale that orders individuals more accurately according to their total scores (Sijtsma & Molenaar, 2002).…”