In large-scale tests, the implementation of computer-based testing (CBT) allows to automatically collect data not only on the students' response accuracy (RA) based on item responses of the test, but also on their response time (RT). RTs can provide a more comprehensive view of a test-taker's performance beyond just what is obtainable based on correct responses alone. In this paper a joint approach is considered to improve the estimation of ability scores involving complex data coming from computer-based test administration. The study focuses on analysing the data of Italian grade 10 mathematics national assessment administered by the National Institute for the Evaluation of the Education and Training System (INVALSI). In addition, a bivariate multilevel regression with speed and ability estimates, obtained by joint model, is developed including individual covariates to evaluate the contribution of individual and contextual variables in predicting test-taking speed and ability. Overall, the main results indicate that mathematics ability and speed are significantly and negatively correlated, and that the hierarchical data structure (students nested into classes) should be taken into account when explaining the dependency of ability and speed on explanatory variables, such as prior achievement, test anxiety, sociodemographic covariates, class compositional variables, school tracks and geographical area.
Bungaro L., Desimoni M., Matteucci M., Mignani S. (In stampa/Attività in corso). Jointly exploring mathematics ability and speed in large-scale computer-based testing. STATISTICAL METHODS & APPLICATIONS, xxx, 1-22 [10.1007/s10260-024-00762-0].
Jointly exploring mathematics ability and speed in large-scale computer-based testing
Bungaro L.;Matteucci M.
;Mignani S.
In corso di stampa
Abstract
In large-scale tests, the implementation of computer-based testing (CBT) allows to automatically collect data not only on the students' response accuracy (RA) based on item responses of the test, but also on their response time (RT). RTs can provide a more comprehensive view of a test-taker's performance beyond just what is obtainable based on correct responses alone. In this paper a joint approach is considered to improve the estimation of ability scores involving complex data coming from computer-based test administration. The study focuses on analysing the data of Italian grade 10 mathematics national assessment administered by the National Institute for the Evaluation of the Education and Training System (INVALSI). In addition, a bivariate multilevel regression with speed and ability estimates, obtained by joint model, is developed including individual covariates to evaluate the contribution of individual and contextual variables in predicting test-taking speed and ability. Overall, the main results indicate that mathematics ability and speed are significantly and negatively correlated, and that the hierarchical data structure (students nested into classes) should be taken into account when explaining the dependency of ability and speed on explanatory variables, such as prior achievement, test anxiety, sociodemographic covariates, class compositional variables, school tracks and geographical area.File | Dimensione | Formato | |
---|---|---|---|
s10260-024-00762-0 onlinefirst 30 luglio 24.pdf
accesso aperto
Descrizione: Versione pubblicata online luglio 2024
Tipo:
Versione (PDF) editoriale
Licenza:
Licenza per Accesso Aperto. Creative Commons Attribuzione (CCBY)
Dimensione
1.02 MB
Formato
Adobe PDF
|
1.02 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.