Type I errors for the 3 Models in the 4 Tuscan samples.

For each of the modern populations listed on the Y axis, data were simulated according to three models and attributed by the LR procedure to one of the models on the X-axis. The power of the procedure in recovering the correct model is represented by the rates of correct attribution (along the main diagonal; shaded cells); the last column (Type I error) represents the fraction of cases in which the correct model was not identified.