Charts

Eight chart functions that know what an actuarial figure should look like, on one colour-vision-validated palette. In Python they draw with matplotlib (pip install "scelo[viz]") and return the figure — no pyplot state, so a notebook shows it once and a script calls fig.savefig(...). In R they draw with base graphics on the current device and return their data invisibly.

PythonR

sc.plot_projection(sc.basicterm(mp))            # cash flows · cumulative PV · in force
sc.plot_relativities(m)                         # forest plot of a GLM's relativities
sc.plot_triangle(tri)                           # development curves, oldest light → latest dark
sc.plot_rates(df, "province", "lapsed", exposure="exposure")   # rate by group, 95 % CI
sc.plot_scr(scr)                                # the SCR build-up with diversification
sc.plot_csm(ifrs)                               # CSM closing balance and yearly release
sc.plot_bars(series)                            # one-series bars, values at the tips
sc.plot_lines(df, "year", ["actual", "expected"])   # up to five series — then it refuses
sc.save_figure(fig, "projection.png")

sc_plot_projection(sc_basicterm(mp))
sc_plot_relativities(m)
sc_plot_triangle(tri)
sc_plot_rates(df, "province", "lapsed", exposure = "exposure")
sc_plot_scr(scr)
sc_plot_csm(ifrs)
sc_plot_bars(x)
sc_plot_lines(df, "year", c("actual", "expected"))

A real plot_projection(sc.basicterm(sc.sample("lifelib-mp"))):

and plot_relativities on the claims-sample GLM from the pricing chapter:

The house style, enforced

The functions carry the lab's chart rules so you do not have to remember them:

• One hue for one series. Magnitude charts use a single green; comparisons use up to five validated series colours — and only the first three pass every pairwise colour-vision check, which is why plot_lines takes at most five series and refuses more instead of drawing an unreadable rainbow.
• Ordinal data gets an ordinal ramp. plot_triangle draws origins oldest-light to latest-dark and labels the first and last.
• Uncertainty is drawn, not implied. plot_rates puts a 95 % Poisson interval on every bar and an overall-rate reference line; plot_relativities puts the interval on every dot and marks the base level hollow at 1.
• The numbers are on the chart. Bars are labelled at the tips (up to 30), the projection marks its break-even, the SCR chart labels its diversification credit.

sc.palette() / sc_palette() return the tokens — surface, inks, grid, the five series colours, a five-step sequential ramp and a diverging triple — for figures of your own that need to sit beside these.