Fit parameters (fit.yaml)¶

The fit.yaml file configures the data, model, priors, and sampler settings.

Example¶

data:
  g:
    file: observation_g.txt
    band: g
    trend: 1
    binsize: 0.00139
  i:
    file: observation_i.txt
    band: i
    trend: 1
    binsize: 0.00139
planets: b
tc_pred: 2460844.98
tc_pred_unc: 0.04
chromatic: true
uniform:
  ror: [0.01, 0.15]
  b: [0, 1]
fixed:
  - period
  - u_star

Data¶

Each dataset is listed under data: with a user-chosen key. Multiple datasets can be specified for simultaneous multi-band fitting.

Parameter	Description	Default
`file`	Path to data file (relative to working directory)	required
`band`	Photometric band name (e.g., `g`, `r`, `i`, `z`)	required
`format`	File format: `generic` or `afphot`	`generic`
`binsize`	Bin size in days (`null` for no binning)	5/1440 (~3.5 min)
`trend`	Polynomial detrending order (`null` for none)	`null`
`spline`	Use spline detrending	`false`
`spline_knots`	Number of spline knots	`5`
`add_bias`	Add a bias (constant) column to the design matrix	`false`
`quadratic`	Include quadratic terms for auxiliary regressors	`false`
`trim_beg`	Trim data within this many days of the start	`null`
`trim_end`	Trim data within this many days of the end	`null`
`clip`	Enable sigma-clipping of outliers	`false`
`clip_nsig`	Sigma threshold for outlier clipping	`7`
`chunk_offset`	Add offset columns for data gaps	`false`
`chunk_thresh`	Gap threshold in days for chunk detection	`0`

Data formats¶

generic (default): A whitespace-delimited text file (.txt) with no header row. The first three columns must be time, flux, flux_error, in that order. Any additional columns are automatically used as covariates in the linear detrending model (e.g., airmass, pixel centroids, sky background). Alternatively, a .csv file with a header row and columns named time, flux, fluxerr can be used.

afphot: A CSV file with columns BJD_TDB, Flux, Err.

Model¶

Parameter	Description	Default
`planets`	Planet letter(s) to fit (must match keys in `sys.yaml`)	required
`tc_pred`	Predicted transit center time [BJD]	estimated from data
`tc_pred_unc`	Uncertainty on predicted transit center [days]	`0.04`
`tc_pred_iso`	Alternative: predicted transit time in ISO format	-
`fit_basis`	Parameterization: `duration` or `mstar/rstar`	`duration`
`chromatic`	Fit radius ratio independently per band	`false`
`include_mean`	Fit a mean flux offset per dataset	`true`
`include_flare`	Include a stellar flare model	`false`
`chromatic_flare`	Fit flare amplitude independently per band	`false`
`include_bump`	Include a Gaussian bump model (e.g., spot crossing)	`false`
`chromatic_bump`	Fit bump amplitude independently per band	`false`

Priors¶

Fixed parameters¶

Parameters listed under fixed are held at their sys.yaml values and not sampled:

fixed:
  - period
  - u_star

Supported: t0, period, ror, b, dur, m_star, r_star, u_star.

Uniform priors¶

Parameters listed under uniform use uniform priors instead of the default Gaussian:

uniform:
  ror: [0.01, 0.15]
  b: [0, 1]

For multi-planet systems, you can specify per-planet bounds:

uniform:
  ror: [[0.03, 0.06], [0.01, 0.04]]

Flare model¶

When include_flare: true, configure the flare parameters:

flare:
  tpeak: 2460925.05        # peak time [BJD]
  tpeak_unc: 0.04
  tpeak_prior: uniform
  fwhm: 0.02               # full width at half max [days]
  fwhm_unc: 0.04
  fwhm_prior: uniform
  ampl: 7.5                # amplitude [ppt]
  ampl_unc: 15
  ampl_prior: uniform

Multiple flares¶

Multiple flares can be specified by using lists for any subset of parameters. The number of flares is determined by the length of tpeak. Any parameter given as a scalar is broadcast to all flares; any given as a list must match the number of flares.

Shared parameters (scalar ampl, fwhm, tpeak_unc broadcast to both flares):

flare:
  tpeak: [2459134.77, 2459134.78]
  tpeak_unc: 0.02            # shared
  tpeak_prior: uniform
  fwhm: 0.02                 # shared
  fwhm_unc: 0.04             # shared
  fwhm_prior: uniform
  ampl: 5                    # shared
  ampl_unc: 10               # shared
  ampl_prior: uniform

Fully independent per-flare parameters:

flare:
  tpeak: [2460211.8493, 2460211.863]
  tpeak_unc: [0.001, 0.005]  # per-flare
  tpeak_prior: uniform
  fwhm: [0.01, 0.03]         # per-flare
  fwhm_unc: [0.02, 0.06]     # per-flare
  fwhm_prior: uniform
  ampl: [40, 15]              # per-flare
  ampl_unc: [80, 30]          # per-flare
  ampl_prior: uniform

The parameters that support per-flare lists are: tpeak, tpeak_unc, fwhm, fwhm_unc, ampl, ampl_unc. The *_prior parameters (tpeak_prior, fwhm_prior, ampl_prior) are always scalar and shared across all flares.

Bump model¶

When include_bump: true, configure the bump (e.g., spot crossing) parameters:

bump:
  tcenter: 2460942.08
  tcenter_unc: 0.04
  tcenter_prior: uniform
  width: 0.01
  width_unc: 0.02
  width_prior: uniform
  ampl: 1.0
  ampl_unc: 5
  ampl_prior: uniform

Multiple bumps¶

Multiple bumps work the same as multiple flares -- use lists for any subset of tcenter, tcenter_unc, width, width_unc, ampl, ampl_unc. Scalars are broadcast; lists must match the number of bumps (determined by tcenter length).

Sampler¶

Parameter	Description	Default
`tune`	Number of tuning (burn-in) steps	`2000`
`draws`	Number of posterior draws	`2000`
`chains`	Number of MCMC chains	`2`
`cores`	Number of CPU cores	`2`
`clobber`	Re-run even if saved results exist	`false`