Dimensionality in data measurement indicates the number of data points assessing a variable. It is crucial for evaluating the Rasch Model, which helps analyze variances in the dataset. Using Winsteps version 4.6.0, eigenvalue outputs showed a total raw variance of 74.3%, with 92.2% of this explained by measures, persons, and items. Despite a 5.8 eigenvalue in the first contrast indicating potential multidimensionality, the constructs remain largely unidimensional. Reliability analysis demonstrated high reliability, exceeding established thresholds for person separation (4.52), person reliability (0.95), item separation (3.26), and Cronbach’s alpha (0.971). The Wright Map visualized the relationship between item difficulties and person abilities, revealing a ceiling effect where participants’ abilities exceeded item challenges. Probability curves further illustrated the correlation between participant abilities and item responses, supporting the scale’s reliability in measuring AI literacy in education. The finalized questionnaire comprises 27 retained items appropriate for assessing this competence.
Source link
