Deciding Whether to Respond: A Latent Class Analysis of Nonresponse on Ontario’s Grade 9 Assessment of Mathematics

Authors

  • Ruth A. Childs Ontario Institute for Studies in Education, University of Toronto http://orcid.org/0000-0003-2687-0955
  • Orlena Broomes Ontario Institute for Studies in Education, University of Toronto
  • Monique B. Herbert Ontario Institute for Studies in Education, University of Toronto

DOI:

https://doi.org/10.11575/ajer.v64i1.56472

Keywords:

Nonresponse, missing data, mathematics, large-scale assessment, latent class analysis, Mots clés, absence de réponse, données manquantes, mathématiques, évaluation à grande échelle, analyse de structure latente

Abstract

This study investigates nonresponse on Ontario’s Grade 9 Assessment of Mathematics – in particular, whether or not students responded to all multiple-choice or all open-response items in two test booklets. Whether students responded to all items of one type (multiple-choice or open-response) by booklet (for the first or second day of testing) was modeled, with and without proportion correct scores by item type as covariates, using latent class analysis. Both a 3-class model without the covariates and a 4-class model with the covariates but without direct effects distinguished among students who responded to all items, students who left both multiple-choice and open-response items blank, and students who left only open-response items blank. The results suggest that deciding to respond to all open-response items is distinct from deciding to respond to all multiple-choice items. Attitudes toward mathematics were also more related to the decision to respond to all open-response items than to the decision to respond to all multiple-choice items.

Cette étude porte sur l’absence de réponse au test de mathématiques pour la 9e année en Ontario –nous cherchions notamment à savoir si les élèves avaient répondu à toutes les questions à choix multiples ou bien à toutes les questions ouvertes dans deux livrets d’examen. Une analyse de structure latente a permis la modélisation du comportement des élèves, à savoir s’ils avaient répondu à tous les items d’un type (questions à choix multiples ou questions ouvertes) dans un livret (lors du premier ou deuxième jour des tests) avec et sans des scores reflétant la proportion de bonnes réponses par type d’items comme covariables. Un modèle de classe 3 sans les covariables ainsi qu’un modèle de classe 4 avec les covariables mais sans effets directs ont tous les deux fait la distinction entre les élèves qui avaient répondu à tous le items, les élèves qui n’avaient ni répondu à certaines questions à choix multiples ni à certaines questions ouvertes et les élèves qui n’avaient pas répondu à certaines questions seulement dans le cas des questions ouvertes. Les attitudes face aux mathématiques ont également joué un plus grand rôle dans la décision de répondre à toutes les questions ouvertes que dans celle de répondre à toutes les questions à choix multiples.

Author Biographies

Ruth A. Childs, Ontario Institute for Studies in Education, University of Toronto

Ruth A. Childs is a Professor and the Ontario Research Chair in Postsecondary Policy and Measurement in the Department of Leadership, Higher and Adult Education at the Ontario Institute for Studies in Education, University of Toronto, Toronto, ON, Canada. Her research investigates the design and equity of large-scale assessments, admissions processes, and other evaluation systems.

Orlena Broomes, Ontario Institute for Studies in Education, University of Toronto

Orlena Broomes is a Senior Business and Information Analyst at the Ontario Ministry of Education. Her research interests include melding sociology and psychometric perspectives, and predictive modeling to investigate the impact of social, health and school factors on student educational outcomes; and aligning research, policy and practice for equity in education.

Monique B. Herbert, Ontario Institute for Studies in Education, University of Toronto

Monique B. Herbert is an Assistant Lecturer in the Department of Psychology at York University, where she teaches research methods and statistics. Her research interests are in the area of educational measurement and outcomes, with a focus on instrument design, program evaluation, missing data patterns and the development of quantitative reasoning skills at the postsecondary level.

Downloads

Published

2018-05-02

How to Cite

Childs, R. A., Broomes, O., & Herbert, M. B. (2018). Deciding Whether to Respond: A Latent Class Analysis of Nonresponse on Ontario’s Grade 9 Assessment of Mathematics. Alberta Journal of Educational Research, 64(1), 70–87. https://doi.org/10.11575/ajer.v64i1.56472