Table 3.

Item-level statistics and inclusion criteria for selected items.

Item statistic	Inclusion criteria for dichotomous items	Inclusion criteria for ordinal items
CTT difficulty. The proportion of examinees answering the item correctly	Items with difficulty between 0.10 and 0.90 were included	Items with <0.90 of the cases scoring on either the highest or lowest answer options were included
CTT discrimination. It refers to the item's capacity to distinguish examinees with high and low ability based on their total score	Item-domain correlation above 0.30 Item-domain correlation with item excluded from domain total score above 0.10 Item-total test score correlation above 0.25	Item-domain correlation above 0.30 Item-domain correlation with item excluded from domain total score above 0.10 Item-total test score correlation above 0.25
Item contribution to internal consistency. Each item contributes to increase or decrease the internal consistency depending on its amount of covariance with other items measuring a common developmental domain	Items that increase Cronbach's Alpha coefficient when included as part of the developmental domain	Items that increase Cronbach's Alpha coefficient when included as part of the developmental domain
Developmental domain internal structure. Items should be associated with the domain they intend to measure. CFA techniques empirically determined this relationship	Items with standardized factor loadings above 0.40 Items with positive standardized factor loadings below 0.40 were reviewed by experts	Items with standardized factor loadings above 0.40 Items with positive standardized factor loadings below 0.40 were reviewed by experts
Relationship with age. Given its association with psychological development, age could be considered an external criterion to identify items that intend to measure development	Categorical regression models that predict the item response based on age and sociodemographic covariates (e.g., gender, preschool enrollment status, mother's education level, if available) Items with a positive regression coefficient for age	Ordinal regression models that predict the item response based on age and sociodemographic covariates (e.g., gender, preschool enrollment status, mother's education level, if available) Items with a positive regression coefficient for age
IRT estimates. These additional item level statistics inform about optimal difficulty or discrimination	1PL and 2PL IRT item difficulty included between −3 and 3 2PL item discrimination above 0.50	GRM item discrimination above 0.50

“1PL” and “2PL” refer to the Rasch and two-parameter logistic IRT models. “CFA” refers to Confirmatory Factor Analysis.