Multiple-indicator latent-growth modelling (MLGM) as a framework for validating longitudinal achievement constructs¶

Theory · All theories
Evidence · 4 claims (2 for, 2 mixed) · 1 study (1 associational), q2 · 0 of 1 report an effect size · 4 claims rest on one study

Description¶

MLGM is a second-order latent growth model, a multivariate extension of the first-order latent growth curve model, in which observed indicators are item clusters (goal scores) rather than total scores. Its stated advantage is that "MLGM can test, instead of making assumptions, whether the same latent construct is measured at each point in time or grade level", and it "accounts for important information of the psychometric properties on the indicators". The article uses MLGM with configural, weak and strong invariance constraints to evaluate whether CAT-based achievement constructs shift across grades or time points.

Design Implications¶

Context¶

Requirements¶

• Multiple measures of the achievement construct along the growth trajectory, with indicators such as goal scores assumed to contain homogeneous content across students

Constraints¶

• Strict factorial invariance is not tested because heterogeneous variance across time is often observed
• For CAT data, item-level indicators are impractical because each student receives different items, forcing the use of item clusters

Target Learners¶

• K-12 students in grades 5-7 taking vertically scaled achievement tests

Target Learning Objectives¶

• Valid measurement of student academic growth in reading and mathematics over grades and time

Claims¶

• Map Cat Longitudinal Construct Invariance Ten States [+M]
• Quadratic Growth Fits Map Better Than Linear [+M]
• MLGM model fit failed for the Indiana mathematics test, the sole exception to invariance support [~W]
• The study's invariance findings may be biased by excluding students who missed a test and by sample heterogeneity [~W]

Related Theories¶

• A multi-rater latent growth curve model that removes unique rater variance before modeling growth

Examples¶

• Test publishers and users should routinely investigate longitudinal achievement construct invariance over time

Key Sources¶

• Wang, S., Jiao, H. and Zhang, L. (2013) 'Validation of longitudinal achievement constructs of vertically scaled computerised adaptive tests: a multiple-indicator, latent-growth modelling approach', Int. J. Quantitative Research in Education, Vol. 1, No. 4, pp.383–407. https://www.nwea.org/research/publication/validation-of-longitudinal-achievement-constructs-of-vertically-scaled-computerised-adaptive-tests-a-multiple-indicator-latent-growth-modelling-approach/. doi:10.1504/ijqre.2013.058307