9.2 Variability within and between cohorts
The final model retains 565 items and employs 18 equate groups. Given the difficulty estimates from that model, we can estimate the D-score and DAZ for each measurement.
Figure 6.2 shows large overlaps in the DAZ distributions between cohorts. This finding suggests that the level of child development is similar in different regions of the world. Some studies display more variability in DAZ than others, which is likely to be related to differences in measurement error, as the number of milestones differs widely.
Observe that we used all cohorts for modelling, which may have made them appear more similar than they are. It would be good if we could verify the apparent similarities in level and variability of child development in different regions by other data that were not part of the modelling.