Update (12/03): Event has been cancelled by organiser.
The Programme for International Student Assessment (PISA) is an international comparative survey of educational achievement of 15-year-olds. The PISA programme investigates and compares the performance of education systems worldwide. Moreover, it assesses knowledge and life skills in reading, mathematical and scientific literacy rather than curricular domains. The first PISA study was conducted in 2000 and Malta participated in the 2009, 2015 and 2018 cycles.
This presentation describes the IRT models and their assumptions, as well as the IRT scaling approach used in PISA 2018. For each domain, a unidimensional multiple-group IRT model was fitted based on the two-parameter logistic model (2PL) for the binary item responses and the generalised partial credit model for the polytomous item responses. These models assume that the response probability to an item depends on the difference between the respondent’s trait level and the difficulty of the item. In addition, the model postulates that for every item, the association between this difference and the response probability depends on an additional item discrimination parameter. Moreover, this presentation describes how these models and methods were applied to the PISA 2018 data to produce the national and international item parameters and the plausible values for each domain.
In addition, PISA 2018 uses student, parent and head of school questionnaire information to generate scale score for a number of latent traits. These scale scores are generated by using latent modelling within the framework of confirmatory factor analysis (CFA). CFA treats items as indicators of unobservable personal characteristics of respondents, such as socio-economic status, teacher support, bullying at school, reading enjoyment, and uses combinations of items to develop a model of the latent construct. After testing and confirming a stable latent construct model, scale scores are estimated to serve as numeric values for the latent variables. These scale score have mean 0 and standard deviation 1, which represents the international average. Moreover, measurement invariance testing is carried out to test cross-country comparability. Across the three instruments (head of school, student and parent questionnaires), the majority of scales reached the metric level of invariance allowing comparisons both across countries and between cycles.