Visualization

Status — Reference

Mirrors drmTMB's Visualization (3 in drmTMB). DRM.jl keeps the base package plotting-dependency-free: these return the data a plot needs — a profile-deviance grid, a correlation summary — so any backend can render them. Drawing is optional via the DRMMakieExt package extension (Makie + AlgebraOfGraphics weakdeps); load a Makie backend and AlgebraOfGraphics to activate drm_figure. Default CI does not draw figures — it only gates the method-less stub.

Plotting data providers

DRM.profile_curve Function

profile_curve(fit, k; npoints = 41, span = 3.0, level = 0.95) -> NamedTuple

1-D profile-likelihood curve for coefficient index k, the data needed for a profile diagnostic plot. At each grid value of θ[k], all nuisance parameters are re-optimised, so the returned curve is a likelihood-ratio profile rather than a Wald/quadratic approximation.

Returns (x, deviance, estimate, cutoff, k, param, coef, level):

• x — grid values for θ[k], spanning θ̂[k] ± span·se[k] and including the MLE exactly;

• deviance — 2(ℓ̂ - ℓ_profile) at each x value;

• cutoff — the χ²₁(level) reference line used by profile intervals.

Requires the fitted objective (fit.nll); the model must be fit through drm.

source

DRM.parameter_surface Function

parameter_surface(fit, k1, k2; npoints = 25, span = 3.0) -> NamedTuple

2-D profile-likelihood surface over two coefficients (global indices k1, k2 into coef(fit)), the data behind drmTMB's plot_parameter_surface. At each grid node the remaining parameters are profiled out (re-optimised), so the surface is the genuine profile deviance 2(ℓ̂ − ℓ_profile(θ_{k1}, θ_{k2})), not a quadratic approximation.

Returns (x, y, z, k1, k2):

• x, y — the grid coordinate vectors for θ[k1], θ[k2] (length npoints), spanning θ̂[k] ± span·se[k].

• z — npoints × npoints matrix of profile deviance (z[i,j] at x[i], y[j]); 0 at the MLE, rising away from it. χ²₂ contours give joint confidence regions.

Requires the fitted objective (fit.nll); the model must be fit through drm.

source

DRM.corpairs_data Function

corpairs_data(fit) -> NamedTuple

Summary of the fitted between-response residual correlation for plotting — the data behind drmTMB's plot_corpairs. Returns (rho = …, constant = …):

• rho — the per-observation ρ12 = tanh(Xρ·β̂_ρ) vector for bivariate models; empty for univariate models (no between-response correlation).

• constant — true when ρ12 does not vary across observations (rho12 ~ 1), in which case a single number rho[1] summarises it; false when it varies (rho12 ~ x) and the full vector / a covariate scatter is the right plot.

source

Drawing layer (optional Makie extension)

DRM.drm_figure Function

drm_figure(data; kind = <inferred>, kwargs...)

Draw a DRM plotting-layer figure from a preparer NamedTuple (profile_curve / parameter_surface / corpairs_data). STUB: the drawing method is provided by the DRMMakieExt package extension, which activates only when a Makie backend and AlgebraOfGraphics are loaded (using CairoMakie, AlgebraOfGraphics / using GLMakie, AlgebraOfGraphics). Without them, calling this throws a MethodError asking you to load the plotting weak dependencies.

Supported kinds:

• :profile ← profile_curve — profile deviance with Florence's Confidence Eye (pale compatibility region where deviance ≤ cutoff, darker outline, hollow point estimate)

• :parameter_surface ← parameter_surface — 2-D profile-deviance heatmap / surface

• :corpairs ← corpairs_data — residual ρ12 summary (constant point or per-observation scatter)

Drawing only — no estimation, no engine computation. Default CI does not load Makie; stub tests assert isempty(methods(drm_figure)) without the weakdeps.

source

DRM.plot_profile Function

plot_profile(fit, k; npoints = 41, span = 3.0, level = 0.95, kwargs...)

Thin drmTMB-named alias: profile_curve then drm_figure (kind = :profile). Requires Makie + AlgebraOfGraphics (via DRMMakieExt).

source

DRM.plot_parameter_surface Function

plot_parameter_surface(fit, k1, k2; npoints = 25, span = 3.0, kwargs...)

Thin drmTMB-named alias: parameter_surface then drm_figure (kind = :parameter_surface). Requires Makie + AlgebraOfGraphics (via DRMMakieExt).

source

DRM.plot_corpairs Function

plot_corpairs(fit; kwargs...)

Thin drmTMB-named alias: corpairs_data then drm_figure (kind = :corpairs). Requires Makie + AlgebraOfGraphics (via DRMMakieExt).

source

To draw locally (not in default CI):

using CairoMakie, AlgebraOfGraphics   # activates DRMMakieExt
pc = profile_curve(fit, 2)
fig = drm_figure(pc; kind = :profile)   # Confidence Eye on the profile interval
# or: plot_profile(fit, 2)