We want to compare the IRT-estimated ability vs. age plots for each language model to make sure that nothing is funky with the fits. (Particularly since Danish and Norwegian comprehension item parameters have shifted distributions compared to the other languages.)
The ability distributions per language greatly overlap.
Correlation of comprehension ability and age: r(16190) = .54, p < .001
## `geom_smooth()` using formula 'y ~ x'
Correlation of production ability and age: r(13668) = .62, p < .001
## `geom_smooth()` using formula 'y ~ x'
Correlation of comprehension ability and production ability: r(13467) = .43, p < .001.