NEWS

---

clusteredMSM 0.1.0 (2026-05-27)

Initial CRAN release

• Initial CRAN release of clusteredMSM.

Highlights

• Nonparametric estimation of population-averaged transition probabilities for clustered or independent multistate process data, with cluster-bootstrap pointwise confidence intervals and simultaneous confidence bands.
• Two-sample Kolmogorov-Smirnov-type test, dispatched automatically when a grouping variable appears on the formula's right-hand side. Three regimes via a design argument: "shared" (multicenter, every cluster carries both groups), "cluster_random" (cluster-randomized trial, stratified cluster bootstrap), and "indep_random" (independent observational comparison, unstratified cluster bootstrap). "auto" infers the regime from the data.
• Native support for non-monotone (recovery) multistate processes via a generic product-integral estimator (prodint_AJ()).
• Self-contained: depends only on survival (which ships with R). Removes the dependency on mstate that limited earlier implementations to progressive models.
• Estimation methodology from Bakoyannis (2021) doi:10.1111/biom.13327; two-sample inference for the cluster-randomized and independent-samples designs follows Bakoyannis and Bandyopadhyay (2022) doi:10.1007/s10463-021-00819-x.

Public API

The package exposes a single user-facing function, patp(), with a formula-based interface modelled after survival::Surv():

# One-sample
patp(msm(Tstart, Tstop, Sstart, Sstop) ~ 1,
     data = mydata, tmat = tmat,
     id = "subj_id", cluster = "site",
     h = 1, j = 2, B = 1000, cband = TRUE)

# Two-sample (estimate + test)
patp(msm(Tstart, Tstop, Sstart, Sstop) ~ treatment,
     data = mydata, tmat = tmat,
     id = "subj_id", cluster = "site",
     h = 1, j = 2, B = 1000)

patp() returns an S3 object of class patp with print() and summary() methods.

Input format

Data are supplied in interval format: one row per mutually-exclusive time interval per subject, with columns for interval start time, end time, starting state, and ending state. Censoring is encoded as Sstart == Sstop on the final row. Within each subject, intervals must be temporally and spatially contiguous; this is enforced by strict validation (validate_intervals()).

Known limitations

• The two-sample Kolmogorov-Smirnov-type test uses unit weight at every time point. Bakoyannis (2021) Section 2.5 recommends the harmonic-mean weight W(t) = prod_p Y_p(t) / sum_p Y_p(t), which downweights regions where one group's at-risk set is small and tames tail instability of the difference estimator. A weighted variant following Bakoyannis (2021) Section 2.5 and Bakoyannis & Bandyopadhyay (2022) is planned for v0.2, alongside the linear and L2 tests.
• Confidence bands use an equal-precision-type construction on the cloglog scale; the Hall-Wellner-type construction from Bakoyannis (2021) Section 2.3 will be added in v0.2.