All simulators
Sharp Regression Discontinuity
Local linear fits versus window means and global polynomials at a scholarship cutoff.
Day 73: Sharp Regression Discontinuityestimated jump
4.27
truth
5.00
Scholarship-style data: treatment switches on at x = 0. Local linear fits within the bandwidth (left and right) estimate the jump; a raw comparison of means inside the window is biased by the slope; global fourth-order polynomials use the whole range and are noisier. The true jump is set by the slider.