Added a symbol argument to all plot.*() methods, allowing study-level
markers to be drawn as rectangles, ellipses, diamonds, triangles, crosses,
plus signs, or stars.
Subgroup plots now distinguish subgroups by point symbol as well as colour
by default. The previous appearance, using rectangles for all subgroups,
can be reproduced by setting symbol = "rectangle".
Added the size = "sampsize_revman" and size = "se_revman" options,
following RevMan-style symbol-sizing conventions, to support visual
consistency with as_revman() exports. The analogous
size = "eb_revman" option is available for latent-class models.
Corrected a bug in plot.Reitsma() with the default size = "fisher".
The covariance used to construct the matrix G for the SROC plot is now
calculated correctly.
Added the size = "fisher" option to plot.RutterGatsonis(),
plot.RutterGatsonisSubgroup(), and plot.ReitsmaSubgroup(). It is now
the default in plot.RutterGatsonis().
Added fitReitsmaTMB() and fitReitsmaSubgroupTMB(), which use custom
TMB templates rather than glmmTMB and may provide improved numerical
robustness in difficult or near-boundary cases.
Corrected a bug in the plot() and as_revman() methods for
ReitsmaSubgroup models when sensitivity and/or specificity are
constrained to be common across subgroups.
Added optional study-level confidence intervals (studyCI = TRUE) and
study labels (studylabels = TRUE) to SROC plotting methods.
Improved starting-value initialization across model families. Robust, finite fallback values are now used when preliminary model fits or empirical estimates are unavailable, including for sparse and degenerate datasets.
Revised the likelihood-ratio-test examples and documentation to caution against using the ordinary chi-squared reference distribution when a reduced model constrains one or more random-effects variances to zero. The corresponding zero-variance boundary comparison was removed from the vignette.
Fixed manual-page titles to use title case consistently.
summary() methods now return objects of dedicated S3 classes
("summary.DTAmodel" and model-specific subclasses) rather than plain
lists, with a corresponding print.summary.DTAmodel() method.
Refactored forest-plot construction across model classes into shared
internal helpers (getForestSensSpec(), getForestSensSpecLCA(), and
getForestPlot()) to reduce code duplication.
coef() and vcov() methods are now available for all model classes.
Examples were updated to use summary() and these accessors rather than
accessing internal list components directly.