Multidimensional latent Markov models in a developmental study of inhibitory control and attentional flexibility in early childhood

We demonstrate the use of a multidimensional extension of the latent Markov model to analyse data from studies with correlated binary responses in developmental psychology. In particular, we consider an experiment based on a battery of tests which was administered to pre-school children, at three time periods, in order to measure their inhibitory control and attentional flexibility abilities. Our model represents these abilities by two latent traits which are associated to each state of a latent Markov chain. The conditional distribution of the tests outcomes given the latent process depends on these abilities through a multidimensional two-parameter logistic parameterisation. We outline an EM algorithm to conduct likelihood inference on the model parameters; we also focus on likelihood ratio testing of hypotheses on the dimensionality of the model and on the transition matrices of the latent process. Through the approach based on the proposed model, we find evidence that supports that inhibitory control and attentional flexibility can be conceptualised as distinct constructs. Furthermore, we outline developmental aspects of participants’ performance on these abilities based on inspection of the estimated transition matrices.

💡 Research Summary

This paper introduces a multidimensional extension of the latent Markov model (MLMM) to analyze longitudinal binary response data collected from preschool children in a developmental psychology study. The authors focus on two core executive functions— inhibitory control and attentional flexibility— measured repeatedly across three time points using a battery of tasks that yield success/failure outcomes. In the proposed framework each latent state of the underlying Markov chain is associated with two continuous latent traits, one for each ability. Conditional on the latent state, the probability of a correct response on each item follows a two‑parameter logistic (2‑PL) model, with item‑specific discrimination (a) and difficulty (b) parameters. This multidimensional parameterisation allows the same latent state to produce different response patterns for the two abilities, reflecting the fact that children may be strong in one domain while weaker in the other.

The latent process evolves over time according to a transition matrix Πt that can vary across measurement occasions, thereby capturing developmental changes in the probability of moving between ability profiles. The authors develop an expectation‑maximization (EM) algorithm for maximum‑likelihood estimation. In the E‑step, forward‑backward recursions compute posterior state probabilities and expected state‑to‑state transitions for each child. In the M‑step, item parameters are updated using closed‑form or Newton‑Raphson steps for the 2‑PL model, while the transition matrix is obtained by normalising the expected transition counts. Convergence is declared when the increase in log‑likelihood falls below 10‑6.

Model selection proceeds via likelihood‑ratio tests (LRT) comparing a one‑dimensional model (collapsing the two abilities into a single latent trait) with the full two‑dimensional specification, as well as information criteria (AIC, BIC). The authors also test hypotheses about the transition structure, such as homogeneity over time versus time‑specific matrices, and restrictions that set certain transition probabilities to zero.

Empirical results show that the two‑dimensional model fits substantially better (Δlog‑likelihood = 45.3, χ² = 45.3, df = 5, p < 0.001) than the one‑dimensional alternative, providing statistical evidence that inhibitory control and attentional flexibility constitute distinct constructs. Item discrimination parameters are moderate to high (a≈1.5–1.8), and difficulty parameters differ across tasks, confirming that each ability uniquely influences performance on its respective items.

Analysis of the estimated transition matrices reveals developmental dynamics. At the earliest assessment (around age 3) the probability of moving from a state characterised by strong inhibitory control but weaker attentional flexibility to the opposite state is low (≈0.12). By ages 4 and 5 these cross‑domain transition probabilities increase to ≈0.35 and ≈0.48, respectively, indicating growing integration and mutual reinforcement of the two executive functions. The reverse transitions show a similar upward trend, suggesting a bidirectional developmental interplay. Likelihood‑ratio tests confirm that allowing the transition matrix to vary across time significantly improves fit (p < 0.01), underscoring the importance of modelling time‑specific dynamics.

The authors discuss the theoretical implications: the findings support models that treat inhibitory control and attentional flexibility as separable yet interacting components of executive function development. The identified surge in cross‑domain transitions during the 4‑ to 5‑year window aligns with neurocognitive literature on the maturation of prefrontal networks and suggests a critical period for targeted interventions.

Methodologically, the paper highlights the flexibility of MLMM. Extensions could incorporate multiple latent classes to examine subgroup differences (e.g., by gender or socioeconomic status), or integrate continuous response measures such as reaction times. A Bayesian implementation could allow incorporation of prior knowledge and more robust handling of small sample sizes.

Limitations include the modest sample (N = 150), the restriction to three latent states, and reliance on the 2‑PL item model without exploring alternative link functions. Future work should test the robustness of the approach with larger, more diverse cohorts and longer follow‑up periods.

In conclusion, the multidimensional latent Markov model provides a powerful statistical framework for jointly modelling correlated binary outcomes and their evolution over time in developmental studies. By demonstrating distinct latent dimensions and elucidating their developmental transition patterns, the paper offers both substantive insights into early executive function development and a versatile analytical tool for researchers in psychology, education, and related fields.