Illustration of how the fitted data reject the best-fit in previous works. The bottom-left panel shows the χ² of the individual data points (from Balestra et al. 2007, chosen for illustration) vs. redshift. The χ² is computed using the Balestra et al. (2007) best-fit. The top left-hand panel shows the same, but for data extracted from a realisation of the Balestra et al. (2007) fitted model. Real data scatter much more than simulated data, as quantified in the top right-hand panel: the total χ² of 30 000 realisations of the Balestra et al. (2007) model is histogrammed, together with the total χ² of their sample (vertical line). Either using χ² tables (or the usual rule χ²/ν, ν ≈ 54) or more sophisticated simulations, the Balestra et al. (2007) data reject their best-fit at > 99.9 confidence.