Fitted SHPM and C_≥1PM-averaged regression lines for the 8 LDTAs. Vertical axis represents expected fraction, termed y_i, of negative mice predicted by the SHPM, or expected fraction, termed y_i, of negative mice predicted by model averaging across the set P of plausible C_≥1PMs; and horizontal axis, number of injected cells x_i at each cell dose group i. (A) LDTA no. 1a: Hoxa-9^−/− bone marrow cells. (B) LDTA no. 1b: wild-type bone marrow cells. (C) LDTA no. 2a: Notch ligand (delta1-IgG) stimulated CD34⁺ cells, 3 weeks after transplantation. (D) LDTA no. 2b: CD34⁺ cultured with control IgG, 3 weeks after transplantation. (E) LDTA no. 2c: noncultured CD34⁺, 3 weeks after transplantation. (F) LDTA no. 2d: Notch ligand (delta1-IgG) stimulated CD34⁺ cells, 9 weeks after transplantation. (G) LDTA no. 2e: CD34⁺ cultured with control IgG, 9 weeks after transplantation. (H) LDTA no. 2f: noncultured CD34⁺, 9 weeks after transplantation. Blue line indicates fitted SHPM regression line; red line, fitted C_≥1PM-averaged regression line; and black symbols, experimental data (y_i/N_i, x_i), where y_i is the number of mice with negative outcome, N_i is the total number of mice, and x_i is the number of injected cells, at each cell dose i. (A-H) The values of χ²/df ratios (Pearson χ²-dispersion statistics) highlight that the C_≥1PM class better fit to the data than the SHPM in 7 of the 8 LDTAs: the lower value of this ratio, the better fit of the model to the data (supplemental data).