Changelog

Source: NEWS.md

circlss 0.1.0

First release.

Circular families for mgcv

Twelve circular-response distributional families written against mgcv’s general-family interface, so gam() fits them directly — each distribution parameter on its own (optionally smooth) predictor, for circular–linear and circular–circular regression:

• von Mises (vmlss()), projected normal (pnlss()), wrapped Cauchy (wclss()), wrapped normal (wnlss()), cardioid (cardlss()), Cartwright (cartlss())
• Jones–Pewsey (jplss()) with its sine-skewed (ssjplss()) and asymmetric (ajplss()) forms, Kato–Jones (kjlss()), flat-topped von Mises (vmftlss()), and inverse Batschelet (ibslss())

Weight-aware Gaussian (gausslss()) and gamma (gammalss()) location-scale families cover a linear response over a circular covariate.

Front-end and helpers

• circ_gam() — thin gam() wrapper for these families: auto-pins cyclic-smooth period knots, fills unmodeled parameters with ~ 1, defaults to method = "REML", and adds geometry-aware print / plot.
• circ_mix() — finite mixtures of circlss families fitted by EM.
• circ_lm() — circular-response linear models.
• Helpers: rad() / deg() / wrap() for radian and branch handling; circ_resid() / circ_check() for diagnostics.