NEWS.md
coef() returning NULL for sma/ma fits: coef.sma() was never registered as an S3 method, so the generic did not dispatch to it for users of the installed package (#26).test/ directory of user bug reproductions; the checks that remain relevant are now covered by the testthat suite (#26).ggplot.sma(), a ggplot2 method for sma objects. Call ggplot() on a fit to draw the data and fitted (S)MA line(s), coloured by group where present, with log-scaled axes for any log-transformed variables (#33). Revives earlier work by Fonti Kar.object 'xi' not found error in slope.com() (and therefore sma() with a grouping variable, e.g. sma(y ~ x * group)) that occurred when the grouping factor contained a level with no observations — for example, an unused level retained after subset(). This also affected sma(..., multcomp = TRUE). Empty groups are now dropped before fitting (which also removes a latent bug where a non-first empty group silently reused the previous group’s variance matrix), and a common-slope test with fewer than two usable groups gives a clear error instead of crashing (#30, #34).