TransitModelFit#

class stellarphot.transit_fitting.TransitModelFit[source]#

Bases: object

Transit model fits to observed light curves.

The underlying transit light curve is computed with pytransit’s RoadRunnerModel using a quadratic limb-darkening law and a circular orbit. The fit is performed with lmfit.

Attributes:

paramslmfit.Parameters: The transit model parameters: t0, period, rp (planet radius in stellar radii), a (orbital radius in stellar radii), inclination (degrees), limb_u1/limb_u2 (quadratic limb-darkening coefficients) and the linear trend coefficients airmass_trend, width_trend and spp_trend. Read or set values, bounds and whether a parameter is fit through the standard lmfit attributes, e.g. mod.params["rp"].value, mod.params["t0"].vary, mod.params["a"].min and, after a fit, mod.params["rp"].stderr.
fit_resultlmfit.minimizer.MinimizerResult or None: Result of the most recent call to fit, including fit statistics like bic and nvarys. None until fit has been run.
times, airmass, width, spp, data, weightsarray-like or None: Independent variables and data for the fit; see the property docstrings.

Notes

The orbit is always circular; there is no eccentricity parameter.

Trend fitting is opt-in: the trend coefficients default to vary=False even when a covariate (airmass, width, sky per pixel) is set. Enable them via TransitModelOptions, directly (e.g. mod.params["airmass_trend"].vary = True) or with compare_detrend_options(apply_best=True).

Examples

Fit a transit, detrending against airmass:

mod = TransitModelFit()
mod.setup_model(t0=2455001.5, depth=12.1, duration=0.15, period=3.5)
mod.times = times
mod.data = fluxes
mod.weights = 1 / flux_errors
mod.airmass = airmass
mod.params["airmass_trend"].vary = True

result = mod.fit()

print(mod.params["rp"].value, mod.params["rp"].stderr)
print(result.bic)

# Which detrending parameters does the BIC favor?
bic_table = mod.compare_detrend_options()

Attributes Summary

`airmass`	airmass : array-like
`data`	data : array-like
`spp`	spp : array-like
`times`	times : array-like
`width`	weights : array-like

Methods Summary

`compare_detrend_options`([apply_best])	Compare, via the Bayesian Information Criterion, fits with every on/off combination of the trend parameters whose covariate is set.
`data_light_curve`([data, detrend_by])	Function to return data light curve, optionally detrended by one or more parameters.
`fit`()	Fit the model to the data and update `params` with the best-fit values and uncertainties.
`model_light_curve`([at_times, detrend_by])	Calculate the light curve corresponding to the model, optionally detrended by one or more parameters.
`setup_model`(*[, binned_data, t0, depth, ...])	Configure a transit model for fitting.

Attributes Documentation

airmass#

airmassarray-like: Airmass at each time. Must be set before fitting.

data#

dataarray-like: Observed fluxes. Must be set before fitting.

spp#

spparray-like: Sky per pixel at each time. Must be set before fitting.

times#

timesarray-like: Times at which the light curve is observed. Must be set before fitting.

width#

weightsarray-like: Weights to use for fitting. If not provided, all weights are set to 1.

Methods Documentation

compare_detrend_options(apply_best=False)[source]#

Compare, via the Bayesian Information Criterion, fits with every on/off combination of the trend parameters whose covariate is set.

Each candidate fit runs on a copy of params — with the trends in the combination varying, and the others fixed at zero — so this method does not change the state of the model.

Parameters:

apply_bestbool, optional: If True, apply the lowest-BIC combination to params (each trend in the winning combination is varied, the others are zeroed and fixed) and run fit.

Returns:

astropy.table.Table: One row per combination with a bool column per available trend (airmass, width, spp) and a BIC column, sorted by ascending BIC (best first).

data_light_curve(data=None, detrend_by=None)[source]#

Function to return data light curve, optionally detrended by one or more parameters.

Parameters:

dataarray-like, optional: Data to use for calculating the light curve. If not provided, the data used for fitting will be used.
detrend_bystr or list of str: Parameter(s) to detrend by. If None, no detrending is done. If 'all', all parameters that are set will be used for detrending.

Returns:

dataarray-like: Data light curve, detrended if requested.

fit()[source]#

Fit the model to the data and update params with the best-fit values and uncertainties.

Returns:

lmfit.minimizer.MinimizerResult: The full fit result, also stored as fit_result. Includes fit statistics like bic.

Raises:

ValueError: If the times or the data have not been set. If the fit itself raises, params and fit_result are left untouched.

model_light_curve(at_times=None, detrend_by=None)[source]#

Calculate the light curve corresponding to the model, optionally detrended by one or more parameters.

Parameters:

at_timesarray-like: Times at which to calculate the model. If not provided, the times used for fitting will be used. Because the airmass/width/spp trend covariates are reused as-is, at_times must have the same length as the model’s times; a mismatched length raises ValueError.
detrend_bystr or list of str: Parameter(s) to detrend by. If None, no detrending is done. If 'all', all parameters that are set will be used for detrending.

Returns:

modelarray-like: Model light curve.

setup_model(*, binned_data=None, t0=0, depth=0, duration=0, period=1, inclination=90, airmass_trend=0.0, width_trend=0.0, spp_trend=0.0, model_options=None)[source]#

Configure a transit model for fitting. The duration and depth are used to estimate underlying fit parameters; they are not themselves fit parameters. Any previous parameter settings are reset to their defaults before the new values are applied.

All arguments are keyword-only.

Parameters:

binned_dataastropy.timeseries.BinnedTimeSeries, optional: Binned time series to load the times, data, weights and covariates from.
t0float: Time of the center of the transit. Can be in any units but should be consistent with the units for the period and for the times used for fitting.
depthfloat: Depth of the transit in parts per thousand. If zero (the default), the planet radius rp is left at its default starting value instead of being estimated.
durationfloat: Duration of the transit, in the same units as t0 and period. If zero (the default), the orbital radius a is left at its default starting value instead of being estimated. Must be smaller than the period.
periodfloat: Period of the planet. Should be in the same units as t0 and times used for fitting.
inclinationfloat: Inclination of the orbit, in degrees.
airmass_trendfloat, optional: Coefficient for a linear trend in airmass.
width_trendfloat: Coefficient for a linear trend in stellar width.
spp_trendfloat: Coefficient for a linear trend in sky per pixel.
model_optionsTransitModelOptions, optional: Options for the transit model fit, mapped onto the vary flags (and, for t0, the bounds) of the parameters.

Returns:

None: Sets values for the model parameters.

Raises:

ValueError: If duration is not smaller than period.