NEXT STEPS IN WRITING

- insert a description of the general approach I use: 
    - predicting, based on snapshots, the likelihood of termination.
    - this needs to go between the description of the snapshots and the 
    causal inference introduction.
- Then I can use what I've written about the graph, and follow up with more information about the data.

Overall this would look like

- [x] Introduction of the question and general issues of confoundedness.
- [x] Clinical Trials Data Sources
- [x] Explain basic econometric modelling approach
- [ ] Then explain the graph, nodes, and confoundedness in more detail
- [ ] Then go over the rest of the data.
- [ ] Finally
    - Discuss the number of datapoints.
    - review major challenges to causal identification. (no enrollment model small data size)