“I’m attempting a reproducible reanalysis of GW150914 post-merger residuals using public data.
Several early papers reference full LALInference posterior files that no longer appear to be accessible.
Can you advise whether these files are still available, or what the collaboration now considers the authoritative public substitute for ML waveform subtraction?”
The frame files used in the residuals analysis in the GWTC-1 testing GR paper are released here. Unfortunately, the corresponding parameter estimation data release (linked, e.g., from the testing GR paper’s data release) does not give enough information to be able to reproduce those frames. However, as far as I remember, the collaboration did not release additional parameter estimation information at that time, which is why the testing GR paper released the frames of the residuals. Is this what you are referring to by “full LALInference posterior files that no longer appear to be accessible”? If not, can you give more details here, so I can investigate if something that should still be accessible is no longer?
However, the analysis in the GWTC-1 testing GR paper was performed with what is now a rather old waveform; the most up-to-date LVK parameter estimation analysis of GW150914 (with at that time some of the latest waveform models) was carried out in the GWTC-2.1 paper, and these improved waveform models do recover a slightly higher SNR. The GWTC-2.1 posterior sample release (linked, e.g., from the latest GW150914 GWOSC page) contains the full information you would need to be able to subtract the maximum likelihood waveform.
Additionally, if you are interested in using the numerical relativity surrogate NRSur7dq4, which should be the most accurate waveform in its calibration region, there are posterior samples released by collaboration members here (though this is not an official LVK collaboration product, and thus has not been subject to the collaboration’s internal review).