The code in the repo deals with the problem of modeling the discrepancies between two sets of observations.
Imagine that you have observations v_1 with Gaussian error e_1, and observation v_2 with Gaussian error e_2.
We model dv = v_1 - v_2 by a mixture model P(dv) = f_out * P(dv|outl ) + (1-f_out) * P(dv|good) where two terms refer to outlier model P(dv|outl) and good population model P(dv|good)
All the modeling is done in the interval -vlim<dv<vlim , i.e. we don't model the distribution outside that interval.
The model for outliers is
The model for good data is:
- Citation information:
If you use the code, please cite this repository and also cite the use of the dynesty nested sampler (Speagle 2020, Koposov+2023)