joineR

Bug fixes

• emUpdate() and emUpdateCR() now issue a warning() on non-convergence instead of silently calling print(), making it possible for callers to catch non-convergence programmatically.
• survst() no longer unconditionally forces the first subject's censoring indicator to 1 before passing data to coxph(). A proper stop() is raised if no events are observed. The rs index vector is now computed correctly via findInterval(), clamping subjects who precede the first event time to risk set 1 rather than producing NA.
• survst() and survstCR() now correctly index loglik[2] (the fitted model log-likelihood) rather than storing the full length-2 vector in $log.like.
• Bootstrap confidence intervals in jointSE() now use quantile() instead of a manual sorted-index lookup, fixing an off-by-one error for small n.boot values.

Improvements

• joint() and jointSE() arguments gpt, lgpt, max.it, and tol now have explicit default values in the function signatures, replacing the previous missing() pattern. Defaults are visible in auto-complete and ?joint.
• jointSE() output now includes a p-value column containing two-sided Wald p-values for all fixed-effect and association parameters. Variance component rows (U_*, Residual) return NA as Wald tests are not appropriate for positive-constrained parameters. Closes #58.
• jointSE() confidence intervals are now computed for all values of n.boot. Previously, runs with fewer than 100 bootstrap samples silently returned 0 for both CI bounds; now the empirical percentile CIs are always returned and a warning() is issued when n.boot < 100 to alert the user that the intervals may be unreliable.
• tidy() and glance() methods are now available for joint objects, providing a tidy broom-style interface. tidy() returns a data frame of parameter estimates (one row per term) and optionally accepts a jointSE() result to add standard errors, Wald statistics, p-values, and confidence intervals. glance() returns a single-row summary of model-level statistics including log-likelihood, AIC, BIC, number of observations, number of subjects, convergence status, and EM iteration count. Closes #59.
• simjoint() no longer prints a progress line via cat() from the internal simdat() function. It now emits a message() at the simjoint() level, which can be suppressed with suppressMessages().
• survival has been moved from Depends to Imports. Surv is re-exported so existing user code does not require changes.