NEWS

---

MSTest 0.1.9 (2026-07-13)

Reproducibility note

• The internal standard-normal generator randSN() now uses arma::randn instead of a Box-Muller transform (faster, and avoids a log(0) edge case). IMPORTANT: this changes the random numbers drawn for a given set.seed(), so simulated series (the simu* functions) and Monte Carlo p-values (LMCLRTest(), MMCLRTest(), DLMCTest(), DLMMCTest(), CHPTest()) differ numerically from earlier versions. Test conclusions and validity are unaffected.

Testing procedures

• MMCLRTest(): the observed likelihood ratio statistic is now held fixed across the nuisance-parameter search, following Dufour (2006), instead of being recomputed at each candidate value. This corrects the maximized Monte Carlo p-value (which could previously be slightly liberal).
• MMCLRTest(): pre-drawn simulation innovations are now held fixed across the optimization (fixed-error Monte Carlo; Dufour 2006, Prop. 4.2).
• LMCLRTest() and MMCLRTest(): added an mc_seed control for fully reproducible Monte Carlo p-values; the number of parallel workers is capped at N when more workers than replications are requested.
• LMCLRTest() and MMCLRTest(): robust handling of failed simulated draws. Non-finite draws are dropped before the p-value is computed. If the simulated null distribution cannot be built at all (every draw fails even after the re-draw safety), the functions now stop with an informative message (increase use_diff_init or inspect the fit) instead of returning an invalid value or crashing the optimizer. In MMCLRTest(), a candidate parameter value whose null cannot be simulated is penalized so the optimizer avoids it; the initial value (theta_0) is exempt so that the MMC p-value remains at least as large as the LMC p-value.
• MMCLRTest(): the GA optimizer now uses popSize = 10 by default (GA's own default of 50 makes the search perform 50 x maxit expensive evaluations); additional GA controls can be passed through optim_control.
• MCpval(): the type argument now accepts both long and short spellings ("geq", "leq", "two-tailed"/"two-tail", "absolute"/"abs") and raises an informative error for an unrecognized value (previously it silently returned a sentinel). The documentation now matches the accepted values.
• DLMMCTest(): fixed the sign of the stationarity penalty so non-stationary candidate parameters are correctly avoided by the optimizer (previously, in some cases the maximized p-value could be returned as the raw penalty constant).
• DLMCTest() and DLMMCTest(): corrected the sample size used to simulate the null distribution of the moment-based statistics. The simulated moments are now computed from samples of length T - p, matching the number of AR(p) residuals used for the observed statistic and for the p-value calibration in approxDistDL() (previously T). This restores the exact exchangeability underlying the Monte Carlo p-value (Dufour & Luger 2017); the effect is negligible for large samples and grows with p/T.
• print()/summary() for DLMMCTest objects now display the nuisance parameter value that maximizes the Monte Carlo p-value (theta_max_min/theta_max_prod) alongside the moment statistics, matching the output of DLMCTest (display only; the returned object is unchanged).

Estimation (EM)

• New conv control for the EM stopping criterion in HMmdl(), MSARmdl(), MSARXmdl(), MSVARmdl(), and MSVARXmdl(), with options "loglik" (relative log-likelihood change; the new default), "theta" (relative parameter change, the previous behavior), "both" (both, following Krolzig 1997), and "loglik-A"/"both-A" (Aitken-accelerated log-likelihood; Böhning et al. 1994, McLachlan and Krishnan 2008). A separate ltol control sets the log-likelihood tolerance (default 1e-7); thtol (default 1e-6) remains the parameter tolerance. NOTE: the default change from parameter-based stopping to "loglik" can change estimates slightly for a given set.seed(); set conv = "theta" to recover the previous behavior. Simulations show "loglik" matches the size and power of "theta" while converging substantially faster. Each fitted model now also reports a converged flag.
• Corrected the M-step mean update for Markov-switching AR and VAR models following Krolzig (1997) (the previous weighted-average update is exact only for hidden Markov models).
• Comprehensive numerical-stability improvements to the EM algorithm: guarded matrix inversions (solve() with fallback), Hamilton-filter underflow handling, degenerate-regime guards, positive-definite covariance regularization, and stationarity checks. Several data sets that previously triggered solve(): solution not found now estimate successfully.
• Faster Hamilton filter and Kim smoother (O(M) forward/backward recursions).
• Correct Kim-smoother handling for the transition-probability update (renormalization fix) and additional guards against transition-matrix corruption in degenerate replications.
• Fixed MSVARXmdl() to call MSVARXmdl_em() (not MSVARmdl_em()) in the single-initial-value EM branch (used when use_diff_init = 1 with a user-supplied init_theta), so exogenous regressors are handled correctly on that path.

Standard errors

• New se_method = "louis" control for the Markov-switching model constructors (HMmdl(), MSARmdl(), MSARXmdl(), MSVARmdl(), MSVARXmdl()), computing expected-complete-data standard errors (Louis 1982); automatic fallback from the observed-information Hessian to Louis when the Hessian is ill-conditioned.
• Transition-matrix standard errors now use a reduced (free-parameter) parameterization, fixing inflated or NA standard errors for k >= 2. Per-parameter step sizes are used in the numerical Hessian to keep transition probabilities within [0, 1].

Other

• Model constructors now validate use_diff_init >= 1 (and maxit_converge >= 1) with a clear error instead of failing cryptically.
• C++ sources: made integer/double literal usage explicit throughout (no change to results).
• Added a unit-test suite (testthat) and an introductory vignette.

MSTest 0.1.8 (2026-01-09)

• made changes to some function examples.

MSTest 0.1.7

• Made change in trycatch error message for HLRTest() to better describe optimization issues.
• Made minor changes to article.R example file.

MSTest 0.1.6 (2025-10-23)

• Updated MLE estimation following deprecation of hin>=0 (inequality constraint direction) in slsqr
• Changed OLS unbiased estimates of models with k=1 to be consistent with MLE estimates. Package is for testing more than estimation so comparison with MLE-based tetsing is prioritized.
• Changed use of arma::is_finite(X) to std::isfinite(X) because former is now deprecated.

MSTest 0.1.5 (2025-02-24)

• patch to compile following (Rcpp) Armadillo update (i.e., added proper namespace scopeas_scalar).

MSTest 0.1.2 (2024-05-31)

• Made change to MSARmdl(), MSVARmdl(), and HMmdl(). Specifically, when msmu or msvar is FALSE, output list copies single regime value for each k regime. This is needed for simulating the null when either msmu or msvar is FALSE in LMCLRTest() and MMCLRTest().
• Added option to use different number of initial values for estimating MSMs with observed data vs. for null distribution (see documentation for LMCLRTest() and MMCLRTest()).
• Updated USGNP data set to include 2022 Q3.
• MMCLRTest() now has the option to add lower and upper bounds for autoregressive coefficients and transition probabilities to help reduce errors from polyroot() when optimizing. MMC_bounds() has been updated to reflect this.
• DLMMCTest() now has the option to add lower and upper bounds for autoregressive coefficients to help reduce errors from polyroot() when optimizing. DLMMC_bounds() has been created for this.
• Fixed error in bootCV() used by CHPTest(). When NaN occurs, new draw is used.
• Added option to allow user to specify 'mle_theta_low' and 'mle_theta_upp' which determine the lower and upper bounds for optimization in HMmdl(), MSARmdl(), and MSVARmdl() when "method='MLE'" is specified.
• print.CHPTest() now used two lines to print description.
• In HLRTest() user can now define entire grid for transition probabilities.
• added new optional optimization routine for HLRparamSearch() (nloptr::slsqp() can be used now).
• Fixed bug in HLRTest() where grid for sigma options are properly used now.
• Now using nearest_spd() from pracma instead of nearPD() from lmf package.
• Added classes for simulation functions
• Added new methods, namely: coef, fitted, predict, summary, residuals, nobs, plot
• Made changes to print method.
• Changed methods for models, namely: logLiklihood now uses logLik, and aic and bic now use AIC and BIC methods.
• Updated USGNP dataset to include data up to end of 2023.
• Updated README.md file with usage of new methods.
• Updated DESCRIPTION file for changed dependencies and new version.

MSTest 0.1.1 (2023-01-31)

• Made changes to MMC LRT related functions for obtaining null distribution of statistical test.
• Added more examples. Examples that take long to complete are commented out but serve to get familiar with usage.
• Fixed bug related to using init_theta (setting initial values of parameters) when estimating Markov models (i.e. MSARmdl(), HMmdl(), and MSVARmdl()).
• Fixed bug in MMC_bound() when k0>1
• Updated DESCRIPTION file for new version.

MSTest 0.1.0 (2022-10-19)

• Added a NEWS.md file to track changes to the package.
• Added a README.md file to describe the package.
• First public version of package.