A MULTICOMPONENT LATENT TRAIT MODEL FOR DIAGNOSIS
SUSAN E. EMBRETSON
GEORGIA INSTITUTE OF TECHNOLOGY
XIANGDONG YANG
EAST CHINA NORMAL UNIVERSITY
A noncompensatory latent trait model, the multicomponent latent trait model for diagnosis (MLTM-D) is introduced in the study. The model is
a generalization of two earlier models, the multicomponent latent trait model (MLTM: Whitely,1980) and the general component latent trait model (GLTM; Embretson, 1984) that is formulated with a hierarchical structure to represent item differences in the number of underlying components. Simulation study was conducted to demonstrate the parameter recovery. Data were generated for the full hierarchical structure of MLTM-D with two attribute design factors specified for each item component.
For the simulation part, there are just results for parameter recovery, no model comparison, how to highlight the strength?
What if there are correlations between the different components?
The model will also be affected by the correctness of Q matrix. It will have the same problem like the usual CDMs.
The model comparison part only using AIC while AIC favors a more complex model, other indices can be added.