NEWS

---

gmwmx2 0.0.6

• Added clear errors when plotting GNSS objects of class gnss_ts_ngl with missing or empty position data.
• Added calendar-year axes to time-series plots of gnss_ts_ngl objects.
• Updated the NGL load-and-plot vignette to use station CERN and skip plotting when downloaded position data are empty.
• Registered the S3 methods for internal autocovariance and model covariance calculations, resolving roxygen2 registration warnings while keeping the generics internal.
• Updated the README to describe Wavelet Moment Regression (WAMORE) and cite the 2026 arxiv preprint.

gmwmx2 0.0.5 (2026-06-10)

• Added a composable stochastic-model interface. Models are now built with wn(), ar1(), rw(), pl(), flicker(), and matern(), and independent components can be combined with +.
• Added gmwm2() for fitting stochastic models directly from a time series by matching empirical and theoretical wavelet variances. The returned gmwm2_fit objects include print and plot methods.
• Reworked gmwmx2() as an S3 generic. It now supports both the GNSS workflow for gnss_ts_ngl objects and a generic regression interface gmwmx2(X, y, model, ...) for arbitrary design matrices and response vectors, including responses with missing observations.
• Changed GNSS model specification from string labels such as stochastic_model = "wn + pl" to model objects such as model = wn() + pl() or model = wn() + flicker(). GNSS fit objects now use class gmwmx2_fit_gnss_ts_ngl and have updated print and plot methods.
• Added simulation utilities through generate() methods for single stochastic models, composite stochastic models, and two-state Markov missingness models created with markov_two_states().
• Added infrastructure for model parameter initialization, constrained parameter transformations, autocovariance evaluation, theoretical wavelet variance computation, and model-implied covariance matrices.
• Added random-walk covariance C++ routines and registered them through RcppExports.
• Expanded documentation and vignettes for data generation, composite stochastic model estimation, generic dependent-error regression, regression with missing observations, and GNSS time-series estimation.
• Updated package metadata for the new modeling stack, including the longmemo dependency and pkgdown website URL.

gmwmx2 0.0.4 (2026-01-07)

• updated functions download_all_stations_ngl, download_estimated_velocities_ngl and download_station_ngl so that it exit gracefully if NGL server is not reachable.

gmwmx2 0.0.3 (2025-08-19)

• Solve minor bug when creating design matrix if jumps or earthquakes are indicated after the last observation.

gmwmx2 0.0.2 (2025-04-10)

• updated functions download_all_stations_ngl, download_estimated_velocities_ngl and download_station_ngl with new address of the NGL: https://geodesy.unr.edu and using httr2.

gmwmx2 0.0.1 (2025-03-10)

first version