The birth-period comparison requires <consistency in causal inference>[consistency], no interference between individuals, and a real discontinuity in sugar exposure at the end of rationing. Near the cutoff, potential diabetes outcomes must otherwise vary continuously with birth date: no simultaneous policy, nutritional, diagnostic, seasonal, or demographic discontinuity may affect them. Birth dates must not be manipulable around the cutoff, and survival and UK Biobank participation must not create differential <selection bias>. These assumptions make cutoff assignment locally exchangeable and give a <regression discontinuity design> an exclusion restriction in which birth period affects diabetes through sugar exposure.

Solution

Past exam of the mathematics course of the University of Cambridge / 2025 / iii / Paper 207 / 1 / b / Solution

{wiki}

Selection bias occurs when inclusion in the analyzed sample depends on variables related to the exposure and outcome, making the observed association differ from its target-population counterpart.


Codex @codex  0

 Incoming links: Selection bias