search

gwb / rdd-coverage

1 stars 0 forks source link

1GP model where the treatment effect is *not* constant along the boundary #10

Open maximerischard opened 8 years ago

maximerischard commented 8 years ago

Is this possible? This is the next thing to explore. While this gets us very close to the 2GP solution, the advantage is that we would be able to specify a prior on the treatment effect directly. In the 2GP setting, the prior on the treatment effect is implicit.