gmvjoint version 0.4.0
- Fairly minor updates
surv.formula
now properly accepts factor variables.
This shortcoming was brought to my attention by J. Piekos (Oxford); so
thanks!
residuals.joint
functions have been amended to actually
work for Cox-Snell (there was an issue with e.g. spline time
specifications).
plot.residuals.joint
now produces a series of plots
with meaningful legends.
ROC
and dynPreds
have been updated to
correctly work with splines, and have been altered slightly.
joint
now accepts the control argument
inits
which allows the user to overwrite initial
conditions. The function boot.joint
has been amended in
line with this along with necessary documentation.
- Added CITATION file with link to CSDA paper.
gmvjoint version 0.3.0
- Fairly extensive re-writes to workhorse functions
joint
(and underlying EMUpdate
); some computational efficiency
has been gained here ‘within’ the EM algorithm.
- Removed
CreateDataMatrices
function, in favour of
simply taking ones from fits used in initial conditions (via
glmmTMB
); this greatly reduces computation times,
particularly for larger models.
- Added a bootstrapping function
boot.joint
: The
approximate SEs created by the model are known to be underestimated, and
so this bootstrapping function seeks to provide an alternative route; or
means of comparison.
- Spruced-up plotting of conditional distributions of the random
effects.
- Updates to S3 methods.
gmvjoint version 0.2.0
- Dynamic predictions added
- ROC/AUC function added
- Function to generate bootstrapped confidence intervals for AUC
added.
- Various S3 methods for joint objects added.
- Relaxation of (quite stringent) model calls.
- Extraction of conditional distribution of random effects.
gmvjoint version 0.1.0