Updates to 1st draft - Githubissues

IRT : set of overlapping items that allows to identify the student scores on a common scale across years

[x] Put years on x axis and % on Y axis (fig1)
[x] Instead of public/private and boys/girls, look at reportcard
[x] Motivate it as robustness checks on the papers that are published :
- effect size in raw scores, standard deviations (not the IRT scores)
- As regression results, control for district Fe and with cluster SE
- then with the robustness
[ ] Correlation matrix across different IRT scores
[x] Explain IRT once at the top
[x] Introduce the regression framework at the top (not when introducing the measures)
[x] Move the scaling after the subsetting
[ ] Example of 15 kids
[x] Andres command troubleshoot : check that the new scores are rank preserving
[x] Andres command troubleshoot : run again on math only (instead of total_mle)

BeaLeydier / Test-Score-Sensitivity