Comparison between the Fermat potential difference posteriors from different LENSTRONOMY model settings. The posterior for a particular setting is obtained by averaging over models that differ in the other model settings, but within a particular model family using the procedure described in Sect. 6.3. Top two rows: correspond to the power-law mass model families, and bottom two rows: correspond to the composite mass model families. The shaded regions illustrate the 68% credible regions. The posteriors are blinded by , where the ‘ref’ subscript refers to one of the compared models. The potential differences are consistent within 1σ between our adopted choices of model settings.