Sensitivity to window choices in local randomization approach (Regression Discontinuity Design)

[Khawar Afreen]

I am confused about my results; are they robust or not? my results are:

Panel I
Diff- in means statistics: -0.273
Fisherian p-value: 0.043

Panel II
Diff- in means statistics: -0.353
Fisherian p-value: 0.015

Panel III
Diff- in means statistics: -0.418
Fisherian p-value: 0.023

What are the criteria to check the robustness?