search

farr / AlignedVersusIsoSpin

A calculation and paper comparing aligned and isotropic black hole spin models.
MIT License
5 stars 2 forks source link

Richard's P&P comments #11

Closed SimonStevenson closed 7 years ago

SimonStevenson commented 7 years ago

I strongly suspect the analysis implicitly (in Eq. B1) uses a simplified model for selection biases. Precessing systems are harder to find (mismatch and luminosity); see for example

Low mass http://adsabs.harvard.edu/abs/2012PhRvD..86f4020B

High mass: http://adsabs.harvard.edu/abs/2010PhRvD..82j4006O

A true likelihood will take the latter effect into account if you're using real PE with precessing waveforms; so you should point out that you have checked the answer won't change with full (synthetic) PE, and you should emphasize the calculation looking forward to the future is using a simplified approximation but (I assume/hope!) won't change much with real PE on a distribution of events. However, it won't account for the former (i.e., search biases towards apparently nonprecessing waveforms) So, make sure to acknowledge both related issues somewhere (appendix B? E?)

 Also, greater detail as regards the future-event population and

uncertainties would be helpful. Based on the paper it looks like by design you simply draw 10 random chi_eff values, and the chi_eff error is of order the observed chi_eff errors as a benchmark. So you should probably make an aside (which I suspect is true/easily validated against past work) that the second assumption is reasonable given random draws from a plausible synthetic population.

Richard O'Shaughnessy

farr commented 7 years ago

I think we implemented the suggestion about more detail on chi_eff uncertainties when closing #10, and I propose to ignore the other comment about selection biases. The papers Richard is pointing to are either for high-mass (100-1000 MSun black holes) or for spinning systems searched for by non-spinning (NSBH!) searches. The existing LIGO searches use an extra dimension in the template bank for chi_eff, so already have the most important spin parameter present. For comparable-mass (i.e. not NSBH) systems, precession will not have such dramatic effects on the waveforms; even weaker effects for short waveforms associated with higher-mass BBH inspirals where there may be only one or two precession cycles in band. Once we've submitted, I'll write something to Richard thanking him for the comment and indicating what we have and have not done on the LIGO P&P website.