search

fbartolic / lens_predict

Predict magnification of ongoing microlensing events with a hierarchical model.
0 stars 0 forks source link

Parameterisation and hierarchical structure of the model #5

Open astrophpeter opened 5 years ago

astrophpeter commented 5 years ago

Which params do we want to learn the population statistics for? Currently, it is just the peak flux change. Do we want to learn the distribution of timescales? Albrow learns the distribution of time since alert to time of maximum, do we want to do this?

fbartolic commented 5 years ago

I'm reformulating the code to include priors on additional parameters. If we only care about peak magnification A(u_0), then we can use a separable prior for say u_0, delta_t_0, and t_E. A better option might be a joint prior (say multivariate Gaussian) on (u_0, t_E), since u_0 is highly covariant with t_E. I'll push some updated code this weekend.