Latent Transition Analysis With Different Measurement Models: Linking Kindergarten Readiness to Early Reading Trajectories

• In Event: New Twists on Latent Transition Analysis: Unique Adaptations and Applications

Abstract

This paper takes a novel approach to latent transition analysis (LTA) by combining different types of latent mixture models - a latent class analysis (LCA) and a growth mixture model (GMM) - rather than the more traditionally used latent class analysis (LCA) of repeated measures. Within this framework, the multinomial logistic regression relates the latent classes of kindergarten readiness to latent classes of reading trajectories. In addition, the three-step method for specifying these complex models while allowing covariates to predict not only who is in each of the classes but how these covariates influences individuals’ probability of transitioning. This novel application of LTA highlights the modeling flexibility that allows different measurement models for the latent class variables.
Using a sample of N = 2,408 Latino(a) students, we collected data on students' kindergarten readiness using the Kindergarten Student Entrance Profile while data on their ELA achievement was collected using scores from the ELA portion of the California Standards Test. Examining the transitions between the two models allowed us to identify which pattern of ELA development a student is most likely to follow given her kindergarten readiness class. Taking the analysis further, when examining transition patterns, we controlled for preschool experience and students' age.
Results showed a five-class solution best fit the LCA data. Classes were mostly differentiated by student responses to items measuring social-emotional and cognitive abilities. Regarding the GMM, a four-class solution was the best fit to the data. The highest class was considered "Above Average" compared to the population of California students. The other three classes all performed below the state average during grades 2 - 5, but varied in their degree of such performance. As may be expected, students who appeared most ready for kindergarten were most likely to transition into the highest-performing pattern of ELA development. Those least ready for kindergarten were also least likely to perform highly in ELA development. Regarding covariates, students who experienced preschool were more likely to be ready for kindergarten, but the effect was not as strong when related to ELA development. However, preschool experience had an indirectly beneficial effect on ELA development in that those who were more ready for kindergarten (as a result of preschool experience) were consequently more likely to perform higher in ELA. Similarly, the effects of age were stronger in regards to kindergarten readiness. Older students were more likely to have higher level of kindergarten readiness, but this effect was less pronounced in terms of ELA development.
This study provides a framework for modeling complex latent processes across time. We move beyond traditional longitudinal approaches limited to repeated measures and connect not only multiple measures, but also multiple types of latent variable models. Practically, when professionals are able to link later developmental processes with earlier functioning levels, necessary interventions may be provided in a more timely manner.

Authors

• Ryan Grimm, University of California - Santa Barbara

• Karen L. Nylund-Gibson, University of California - Santa Barbara

• Matthew Pierce Quirk, University of California - Santa Barbara

• Michael James Furlong, University of California - Santa Barbara