Regresar

Normative Data for Learning and Memory Test (TAMV-I) in Latin American and Spanish Children: An item response theory and linear mixed models approach

Abstract:

Robust normative data for pediatric learning and memory tests in Spanish-speaking populations are scarce, and existing approaches often rely on univariate methods that overlook item-level properties and inter-trial dependencies. The aim was to evaluate the item parameters of the TAMV-I using Item Response Theory (IRT) and to generate covariate-adjusted normative data through Linear Mixed Models (LMM). We hypothesized that the 2-parameter logistic (2PL) model would outperform the Rasch model and that demographic and contextual factors would show significant interactions influencing test performance. The sample consists of 1640 participants from Spain, Honduras, Ecuador, and Colombia. The inclusion criteria were being 6–17 years old, IQ ≥ 80 on TONI-2, and score<19 on the Children’s Depression Inventory (CDI). Children with a history of neurological and/or psychiatric disorders were excluded. Item parameters were determined using the 1,2-PL model. LMM were used to evaluate the effect of sociodemographic variables (sex, age, age<sup>2,</sup> mean parent years of education-MPE, country, and interactions). Norms were generated based on participant ability. As a result, the item parameters were calculated and the LMM showed significant interactions for MPE ∗ Country , Age ∗ Trial, Sex ∗ Country and Trial ∗ Country. By integrating IRT with LMM, this study provides cross-national, covariate-adjusted norms for the TAMV-I, enhancing precision and clinical validity compared to previous approaches.