A claim of "cellular damage from UFO radiation" is only meaningful once it is compared against two things a real radiobiologist would check first: what dose-response the biology actually predicts, and whether the observed damage rate could just be normal background noise. This simulator does both.
S(D) = e^(−αD − βD²) linear-quadratic survival model
p_exposed = p_bg + (1 − S(D))·(1 − p_bg)
z = (p̂₁ − p̂₂) / sqrt(p̂·(1−p̂)·(1/n₁ + 1/n₂))
The linear-quadratic (LQ) model is the standard curve radiobiologists fit to real cell-survival data: the linear α-term dominates at low dose (single radiation tracks causing unrepairable double-strand breaks), while the quadratic β-term dominates at higher dose (two independent tracks combining to break both DNA strands). Together they predict the survival fraction S(D) at any claimed dose, and 1 − S(D) is the dose-induced damage probability layered on top of the cells' baseline (background) mutation rate.
- Claimed dose — the exposure the UFO account asserts; higher doses push more of the LQ curve's damage onto the exposed group.
- α / β — how radiosensitive the claimed cell line is; real human cells span roughly α≈0.1–0.4/Gy, β≈0.01–0.05/Gy² depending on tissue and repair capacity.
- Background mutation rate — the spontaneous chromosome-aberration/mutation rate unexposed cells show anyway, from ordinary metabolic and replication stress. This is the null hypothesis the claim has to beat.
- Cells assayed — sample size per group. Statistical power to detect a real effect (or correctly reject a fake one) rises with n; small assays produce noisy rates that swing past the significance threshold by chance alone.
- Resample assay — redraws both groups from the underlying binomial distributions, so you can see how much the z-score and verdict wobble run to run at a given sample size — exactly the sampling noise a real study has to out-run with replication.
The verdict follows the standard two-proportion z-test thresholds: |z| below 1.96 is not statistically significant (p>0.05, indistinguishable from background); above 1.96 is significant at p<0.05; above 2.58 is significant at p<0.01. A large claimed dose with a tiny sample size routinely fails to clear significance — which is the actual, boring, statistically rigorous reason most "anomalous cellular damage" reports collapse under scrutiny.