Determine whether smoothing can be combined with estimation of transition-noise variance

Determine whether the smoothing extension of the dynamical non-compensatory MIRT model can be effectively combined with estimation of the skill-transition noise variance parameter gamma_k, including appropriate choices of the smoothing weights.

Background

The model includes a smoothing extension that uses nearby response likelihoods to improve latent-skill inference, as well as an option to estimate the transition-noise variance parameter gamma_k. These two options were not combined in the reported experiments because preliminary results were unsatisfactory. The authors explicitly leave unresolved whether the poor preliminary performance reflects a fundamental incompatibility or instead results from inadequate exploration of the smoothing weights.

References

Third, we did not combine the smoothing extension and the estimation of $\gamma_k$ because preliminary experiments did not perform well. This is a current limitation; however, we consider that further exploration of weight parameters is needed to confirm if this combination does not truly work or not.

— Dynamical Non-compensatory Multidimensional IRT Model Using Variational Approximation  (2609.10028 - Tamano et al., 9 Sep 2026) in Discussion, final paragraph of the section on limitations and possible avenues for future research