GTC transmission spectroscopy analysis

1. Overview

Hannu Parviainen, Instituto de Astrofísica de Canarias
Last updated: 22.06.2017

Running the MCMC

The notebooks in this directory are used mainly for the visualisation of the results obtained with the $W80ROOT/bin/fit script.

The fit script can be ran either as

fit --passband PB --nights NN --approach APP --n-threads 1 --n-walkers 800 --de-n-iterations 2500 --de-update-interval 100 --de-save-interval 100 --mc-n-iterations 15000 --mc-update-interval 200 --mc-save-interval 1000 --mc-thin 200 --mc-n-runs 5 --do-mc --use-ldtk

or (using bash)

for i in {1..5}; fit --passband PB --nights NN --approach APP --n-threads 1 --n-walkers 800 --de-n-iterations 2500 --de-update-interval 100 --de-save-interval 100 --mc-n-iterations 15000 --mc-update-interval 200 --mc-save-interval 1000 --mc-thin 200 --do-mc --use-ldtk; done

where PB is the passband in [nb, K, Na]; NN chooses the nights to use from [1, 2, 12]; and APP chooces the modeling approach from [abs, dw, dwr]. The first approach should work fine in theory, but there appears to be a memory leak in the code somewhere, which will eventually lead to a crash. The second approach is better until I find and fix the leak.

Modelling approaches

The code currently implements three modelling approaches

  • ABS: Direct model with GP systematics
  • DW: Divide-by-white model with parametric systematics
  • DWR: Divide-by-white model with GP systematics

These approaches are implemented as classes in modules lpfsd, lpfsr, lpfsrr for single nights, and lpfmd, lpfmr, and lpfmrr for multiple nights. The LPFSD is a subclass of LPF, and LPFSR and LPFSRR are subclasses of LPFSD.


© 2017 Hannu Parviainen