Rodrigues et al. reply—The correspondences by McGaugh et al. 1 and Kroupa et al. 2 question our results 3 (hereafter R18). In essence, they state that our results are in conflict with Li et al. 4 (hereafter L18) and criticize the priors that we used in our analysis. We show that L18 has no implication for our results and that our priors are adequate for our analysis. L18 shows that the radial acceleration relation (RAR) 5 can be inferred from individual fits of galaxies, and they find no evidence for a variable acceleration scale a0. This does not contradict the fact that R18 does find strong evidence against a fundamental a0. The test performed by R18 is more robust and sensitive to this question. Indeed, the test by L18 is based on a visual comparison of the cumulative distribution functions (CDF) of χ2 ν (reduced χ2).

The idea is to see how the overall fitting performance is improved when a0 becomes a free parameter. However, this test cannot infer the compatibility among the a0 values from different galaxies, and it is not fully correct to classify this approach as a Bayesian statistical analysis, as done by McGaugh et al. 1. The definitions of χ2 and χ2