A minimum person separation index of 0.70 and 0.85 is required for group and individual use respectively (Tennant and Conaghan 2007). Rasch analysis also enables investigation of difficulty that clinical educators may have in discriminating between different levels on the 0–4 rating scale. For a good fit to the model it is expected that for any item, student with high levels of the attribute (professional competence
indicated by total scores) would typically achieve a higher item score than individuals with low levels of the attribute. In Rasch Selleck Paclitaxel analysis this is demonstrated by an ordered set of response thresholds for each item. Ordered thresholds indicate that the respondents (ie, clinical educators) use the response categories (ie, scoring scale) in a manner consistent with
the level of the trait (ie, competence) being measured. This occurs when the educators consistently discriminate between response options in a predictable way. A total of 644 APP assessments from SCH 900776 molecular weight 456 students were returned by 298 clinical educators. Tables 1 and 2 present the characteristics of the participating students and educators. Table 3 presents the characteristics of the APP forms received. The mean APP total score was 61 (SD 12, range 16–80). If converted to the 0–100 scale, this equates to a mean total score of 76 (SD 15, range 20–100). All 5 points on the rating scale were used for the majority of items. Missing data was rare (0.4% of all data points) and 0.2% of all items were rated as not assessed. Data were randomly divided into two samples. Sample 1 was used for model development (n = 326) and sample 2 for model
validation (n = 318). The data were stratified before randomisation to optimise representation these of completed APP instruments according to clinical area of the placement, level of student experience, facility type (hospital, non-government agency, community health centre, private practice), and university program type (undergraduate, graduate entry). Overall model fit: The item-trait interaction chi-square statistic for Sample 1 was 65.1 (df = 80, p = 0.88) and 100 (df = 80, p = 0.57) for Sample 2. The chi-square probability values for Sample 1 (p = 0.88) a nd Sample 2 (p = 0.57) indicated adequate fit between the data and the model. Overall item and person fit: The residual mean value for items for Sample 1 was −0.33 (SD 1.71), and for Sample 2 was −0.32 (SD 1.73), indicating some misfit of items. The residual mean value for persons for Sample 1 was −0.26 (SD 1.19) and for Sample 2 was −0.19 (SD 1.13), indicating no misfit of persons in either sample. Individual item and person fit: In both samples, Item 6 (Demonstrates clear and accurate written documentation) exhibited a positive item fit residual above +2.5, suggesting poor discrimination.