virgil.plotting

Grid result with one axis per coordinate key of grid (every key except the flux), in key order, as returned by virgil.grid_fit and virgil.limits. A full likelihood_grid (with the flux axis) is reduced to its maximum over flux. One coordinate gives a line plot.

The grid axes used to compute values.

What values holds, which sets the default colour map, scale and labels: a log likelihood, a best-fit flux, a flux uncertainty, a signal-to-noise ratio, or a flux upper limit.

For kind="flux" and "limit", show the values as flux ratios (companion/primary, default), contrasts (primary/companion, e.g. 100), or magnitude differences (e.g. 5 mag).

Confidence of a kind="limit" map, for its title (e.g. sigma=5 gives "5σ limit").

Confidence of a kind="limit" map, for its title (e.g. sigma=5 gives "5σ limit").

Coordinates to mark with a cross (truth) or circle (best fit). A dict is looked up by key; a sequence is read in coordinate-key order.

Coordinates to mark with a cross (truth) or circle (best fit). A dict is looked up by key; a sequence is read in coordinate-key order.

Mark the primary at the origin of sky maps (default True).

Key of the flux axis, if it is not the one key ending in flux.

Axes to draw on; a new figure is made if omitted.

Override the defaults for kind: colour map, logarithmic colour scale, colour-bar label and title.

Override the defaults for kind: colour map, logarithmic colour scale, colour-bar label and title.

Override the defaults for kind: colour map, logarithmic colour scale, colour-bar label and title.

Override the defaults for kind: colour map, logarithmic colour scale, colour-bar label and title.

Size of a new figure.

(fig, ax).

A limit map (companion/primary flux ratios, with one axis per coordinate key of grid in key order, as from absil_limits or ruffio_upperlimit), or a radial_profile of one.

The grid axes of a limit map; it must contain dra and ddec axes (also as paths such as "comp.dra").

Show limits as magnitude differences (default), contrasts (primary/companion) or flux ratios (companion/primary). Deeper limits are always drawn lower down.

Legend label; by default built from sigma or percentile (e.g. "5σ limit"), or "Median limit".

Legend label; by default built from sigma or percentile (e.g. "5σ limit"), or "Median limit".

Legend label; by default built from sigma or percentile (e.g. "5σ limit"), or "Median limit".

Shade the 16–84 percentile range of each annulus (default True).

(dra, ddec, flux) of a detected companion to mark.

Annuli, passed to radial_profile.

Annuli, passed to radial_profile.

Annuli, passed to radial_profile.

Axes to draw on, e.g. to overlay several curves.

Line and band colour.

Size of a new figure.

(fig, ax).

Simulated searches, from injection_recovery.

"delta_chi2" (default), "log_bayes_factor" or "max_snr".

The statistic of the real data. It is drawn as a vertical line, with its empirical false-alarm probability (DetectionMC.false_alarm_probability) and the 95% interval of that probability as a point with error bars.

A false-alarm probability to draw as a horizontal line, with the empirical threshold that reaches it (DetectionMC.threshold) as a vertical line.

Draw what the statistic would do at one position fixed in advance (default True): ½ P(χ²₁ ≥ x) for delta_chi2 (Wilks's theorem with the flux on its boundary at zero), the Gaussian tail for max_snr. The gap between this curve and the empirical one is the look-elsewhere effect. log_bayes_factor has no reference.

Axes to draw on; a new figure is made if omitted.

Title; by default it names the statistic and the number of draws.

Size of a new figure. The legend is drawn outside the axes; a tight bounding box (as Jupyter uses) includes it.

(fig, ax).

Simulated searches, from injection_recovery.

"delta_chi2", "log_bayes_factor" or "max_snr", or several of them to compare.

Only the injections of this flux, or in this range (lo, hi), as for DetectionMC.roc. A list gives one curve per entry. By default, every injection.

Only the injections with separations in [lo, hi) (mas); a list gives one curve per bin.

Axes to draw on; a new figure is made if omitted.

Logarithmic false-positive axis (default), from half a null draw's worth (0.5 / n_null) to 1, since detections are claimed at small false-alarm probabilities. On it the chance curve TPR = FPR is a curve, not a straight line.

Local significances (default 3σ and 5σ) to mark. A vertical line shows the false-alarm probability that Wilks's theorem gives them at one position fixed in advance, the one-sided Gaussian tail (0.135% at 3σ, 2.9×10⁻⁷ at 5σ), when it lies on the axis. A circle on every delta_chi2 or max_snr curve shows where the threshold equals that local significance (delta_chi2 = 9 or max_snr = 3 at 3σ): its horizontal distance from the line is the look-elsewhere effect of searching a grid. None or () marks nothing.

Title; by default it names what all the curves share (one statistic, one flux or separation selection), and the legend names what differs.

Size of a new figure. The legend is drawn outside the axes; a tight bounding box (as Jupyter uses) includes it.

(fig, ax).

Simulated searches with injections, from injection_recovery.

The detection statistic, e.g. "delta_chi2".

False-alarm probability of the detection threshold, e.g. 1.35e-3 (the one-sided Gaussian tail at 3σ).

The y-axis, as in plot_contrast_curve: magnitude difference (default, linear), contrast (primary/companion, log) or flux ratio (companion/primary, log). Fainter companions are always drawn lower down, so limit curves from plot_contrast_curve can be drawn on the same axes.

Completeness levels to draw as lines (default 50% and 90%); () draws none.

Bin edges (mas, and flux), as for DetectionMC.completeness. By default every distinct injected separation and flux is a cell.

Bin edges (mas, and flux), as for DetectionMC.completeness. By default every distinct injected separation and flux is a cell.

Axes to draw on; a new figure is made if omitted.

Colour map of the completeness, from 0 to 1.

Add a colour bar (default True).

Title; by default it gives the statistic and the FAP.

Size of a new figure.

(fig, ax).

One orbit, or many as a single KeplerOrbit whose elements are 1-D arrays of equal length (e.g. built from posterior samples).

Measured positions to overlay, with n_sigma error ellipses.

A reference orbit (e.g. the truth of a simulation), drawn as a thicker contrasting line.

Points along each orbit.

Size of the error ellipses, in standard deviations.

Colours of the ensemble and of the reference orbit.

Colours of the ensemble and of the reference orbit.

Opacity of each ensemble line; by default it falls with the number of orbits, so that the density of lines reads as probability.

Colour map for the epochs' times.

Axes to draw on.

Size of a new figure.

(fig, ax).

Model to render.

Width of the field of view in milliarcseconds.

Number of pixels on a side.

Axes to draw on; a new figure is made if omitted.

Axes title.

Quantile (e.g. 0.99) at which to saturate the colour scale, so that faint structure next to a bright star is visible.

Matplotlib colour map.

The data's resolution, from imaging.beam, drawn as a shaded FWHM ellipse in the lower-left corner, as is usual on reconstructed images.

If True, show the image convolved with beam (see imaging.convolve_beam): what the data resolve, rather than the super-resolved image.

The axes drawn on.

E.g. reconstruction.render(npix, fov) - truth.render(npix, fov), in the orientation of render.

Width of the field of view in milliarcseconds.

Uncertainty of the residual, per pixel or overall; if given, the map shows z-scores.

Axes to draw on; a new figure is made if omitted.

Axes title.

A diverging Matplotlib colour map.

The axes drawn on (with a colour bar).

Data to show. Flagged samples are drawn too: the coverage is that of the observation.

Axes to draw on; a new figure is made if omitted.

Colour map of the wavelength, used when the data have more than one wavelength (one colour otherwise).

Size of a new figure.

(fig, ax).

Data to show. Flagged samples are left out. Projected observables (vis_mat/phi_mat) have no baseline and are not shown against it.

Figure size.

(fig, (ax_uv, ax_vis, ax_phi)).

Observed data container.

Mapping label -> prediction summary where each summary contains vis_mean, vis_std, phi_mean, phi_std arrays.

Matplotlib color list. If omitted, cycle C0, C1, ...

Figure size.

Title for the phase panel.

If True, enforce square panel boxes for both subplots.

Axis label for the visibility observable. By default it follows oidata.vis_mode: V² and amplitudes are shown in percent, log-amplitudes and projected (e.g. DISCO) observables as plain values.

(fig, (ax1, ax2)) for visibility and phase axes. Data with no phase observables (e.g. AMIGO DISCOs, which hold every observable in vis) leave the phase axis hidden and give the visibility panel the whole figure.

Mapping label -> pandas.DataFrame containing chain samples.

Columns used for contour/corner plotting.

Truth mapping for columns displayed. When omitted, no truth overlay is drawn.

Per-chain colors.

Columns to show in walk plots. Defaults to columns.

(consumer, corner_fig, walks_fig): the configured ChainConsumer instance and the two figures it drew.

Posterior table for vanilla HMC.

Posterior table for Fisher-reparameterized HMC.

Truth mapping in Cartesian coordinates.

Colors used for HMC and Fisher-HMC chains.

Mapping with the ChainConsumer objects ("cartesian", "polar"), their figures ("cartesian_figures", "polar_figures", each (corner, walks)) and the truth mappings.

Mapping of sample arrays.

Key for right-ascension offsets.

Key for declination offsets.

Key for flux or log10-flux samples.

If True, exponentiate flux_key values as base-10.

Table with dra, ddec, flux, sep, and pa columns.

Mapping with dra, ddec, and flux values.

Cartesian and polar truth dictionaries.