I am seeking some resources or specific equation specifications for whether or not it is possible to specify comparative interrupted time series models or difference and difference models with some student achievement data that I have been using. More specifically, modeling multiple cohorts simultaneously.
Administrators are increasingly interested in the potential effect of teacher-level interventions on student achievement outcomes. A potential issue that I regularly run into is that the interventions created usually target elementary school teachers for example. As a result I end up with data with teachers exposed to the intervention, but who teach different grades. For example, I might have 20 kindergarten, 20 1st grade and 20 2nd grade teachers. It gets more complex at the student level because a child who had a teacher who was exposed to the intervention could feasibly move into classrooms in 1st and 2nd grade where that teacher also had exposure to the intervention.
My thought was that in some ways modeling the student achievement data, having access to multiple cohorts at different grades potentially mirrors an accelerated longitudinal design.
My questions would be:
- Are there any resources where these kinds of quasi-experimental designs are used with accelerated longitudinal data? I have not found any in my online searchers.
- Can you model different cohorts into the parameter estimation of models like comparative interrupted time series models?
- If so, what would those equations look like?
- Having not found a way to specify these different cohorts into one model I often propose many different models: In these kinds of quasi experimental models is it best to always set up different models for (as is in my case) kindergarten teachers who were exposed to the intervention vs. a matched comparative group of kindergarten teachers who were not exposed to the intervention, and then to do the same with 1st grade teachers and then 2nd grade teachers?
