Finding significantly good or bad years for school's pass/fail ratio with changing class sizes I have yearly pass/fail ratios for 3 different schools, spanning 10 years. I want to test which years were particularly good or bad for each school. I also want to test whether any student population characteristics can account for the good or bad years. 
The data is recorded at the individual student level like this:
School Student Year Age SAT Pass/Fail
1      1       2000 18  1300   P
1      2       2000 17  1270   F
2      3       2000 19  1200   P  

What kind of analysis makes the most sense for this context?
 A: My approach would be to leverage some variant of hierarchical modeling. This seems like a natural approach given that students are nested within schools. The hierarchical structure would have an advantage over GLMs (treating the schools as factors in ANOVA) in that it would shrink the errors substantially. A good, education-centric introduction is Judith Singer's pdf Using SAS PROC MIXED to Fit Multilevel Models, Hierarchical Models, and Individual Growth Models. Forget that it has SAS in the title, it's just a great, brief intro to this class of HLMs, available here: 
https://www.ida.liu.se/~732G34/info/singer.pdf 
The difference between your model and hers is that your dependent variable is 0,1 for Pass/Fail where her DVs are grades and test scores (for the most part). In addition, depending on how "rare" failing students are, you may need to integrate zero-inflated considerations into your model. At this point, the issues do become software specific in terms of how you deal with these challenges.
