Source data and statistical results of the 1000-permutation linear correlation test under the Constrained, Lost, and Mixed Model.

This model simulates the most realistic scenario by introducing “fake combinations” (Type-I error noise) that are not derived from pathogenic genotypes but are output by the CGCP method due to limited control group size. The simulation used the command python convergence_loss_times_origion.py.txt 1000 20 20 30 10 1, incorporating a 30% loss rate and a 10% mix rate of fake combinations. The first sheet detailed the first permutation data (from 0.origin.txt and 0.txt), and the subsequent sheets present the final statistical results across 1000 permutations (from result.txt). While the introduction of noise makes this model deviate from the ideal linear formula, it provides a crucial comparison with previous models to evaluate the robustness of the correlation between genotype frequency and disease prevalence in real-world settings. (XLSX)