Skip to contents

Changelog

Source: NEWS.md

exuber 2.0.0

New methodologies from the docs/ research programme, each independently validated against a published number (formula-exact check, table lookup, or a direct Monte Carlo reproduction of the source paper’s own theorem) — see docs/README.md for the full record of what was checked and how.

Critical values

  • Default critical values now come from a shared precomputed store instead of the bundled radf_crit dataset: lag 0-4, every n from the smallest the PSY window allows up to 4000, 2000 replications each. radf() followed by summary()/datestamp()/autoplot() fetches the one (n, lag) table it needs on first use and caches it on disk (tools::R_user_dir("exuber", "cache")), so a lagged specification no longer requires simulating your own cv. The tables are nested (every n of a lag from the same seeded paths); values differ from the old bundled ones by Monte Carlo noise. Needs network access the first time a given (n, lag) is used; radf_mc_cv()/radf_wb_cv() remain the offline route.
  • radf_crit (the bundled lag-0, n <= 600 table) is removed, along with its print.crit method and data-raw/sim-crit.R; the simulation now lives in the sibling exubercrit repository.

Volatility-robust tests

Dating and root inference

  • dating_pdc() — PDC/KS sequential sample-splitting dating, plus type = "wls" for Kurozumi & Skrobotov (2023)’s time-varying-volatility correction.
  • radf_recovery()/radf_recovery_cv() — Phillips & Shi (2014) reverse-regression crisis-origination/recovery dating. Caveat: f_c and the overall false-detection rate are exploratory pending further validation, flagged at call time and in ?radf_recovery.
  • dating_hls() — Harvey, Leybourne & Sollis (2017) SSR/BIC single-bubble dating.
  • dating_hlw() — Harvey, Leybourne & Whitehouse (2020) SSR/BIC multi-bubble two-step wrapper.
  • dating_knp() — Kejriwal, Nguyen & Perron (2025) bias-corrected dating.

Real-time monitoring

  • monitor() — Phillips & Shi (2020) training/monitoring orchestration (Family A), plus Kurozumi (2020) closed-form SADF/ GSADF_s0 boundaries and Homm & Breitung (2012)’s FLUC boundary.
  • monitor_cusum() — Homm & Breitung (2012) CUSUM monitoring, plus Astill et al. (2023)’s volatility-robust CUSUMV kernel variant and HB’s finite-sample boundary.
  • lbi_test()/monitor_lbi() — Breitung & Diegel (2025) static LBI test and its sequential mCUSUM/wCUSUM extension.

Multivariate / panel tests

Alternative paradigms

Naming

  • 12 of the functions above (cobubble_test, contagion_reg, monitor_cusum, dating_hls, dating_hlw, dating_knp, lbi_test, monitor_lbi, dating_pdc, monitor_quantile, quantile_test, ssu_test) were named radf_* in earlier development snapshots of this unreleased version; renamed before release since none of them are actually recursive-ADF-based tests. No deprecated aliases were kept, as the old names never shipped in a CRAN release.
  • radf_monitor() renamed to monitor() (via a brief intermediate monitor_radf(), never released), the flagship of the real-time monitoring family alongside monitor_cusum()/monitor_lbi()/ monitor_quantile() — deliberately carrying no radf/sadf token so it can’t be mistaken for a radf_*() variant, ADF-family internals notwithstanding — see vignette("naming-and-analysis").
  • exuber_functions() added: a queryable registry of every exported function’s family (adf, test, dating, monitor, root, regression), so “what monitoring functions exist” is an actual function call, not a naming convention to memorize.
  • radf_wb_cv2()/radf_wb_distr2() renamed to radf_wb_ps_cv()/ radf_wb_ps_distr() – the 2 suffix named nothing (just “the second wild bootstrap added”); _ps identifies it as Phillips & Shi (2020)’s wild bootstrap (fits a null AR model, resamples residuals, supports a tb training-window boundary), as opposed to radf_wb_cv()’s Harvey et al. (2016) non-parametric multiplier bootstrap, matching this file’s own internal naming (radf_wb_dgp_ps/radf_wb_ps vs. radf_wb_dgp_hlst/radf_wb_hlst) and the package-wide radf_<method>_<qualifier>_cv pattern (radf_sign_dm_cv()). Unlike the renames above, radf_wb_cv2() shipped in the 1.0.0 release (the JSS paper), so this one keeps radf_wb_cv2()/radf_wb_distr2() as deprecated aliases (.Deprecated(), warn-and-forward, see ?exuber-deprecated) rather than a clean break.
  • radf_tt_cv(), radf_sign_cv(), and radf_sign_dm_cv() now all compute badf_cv/bsadf_cv (a time-varying boundary), not just the three scalar critical values — radf_tt()/radf_sign()/ radf_sign_dm() results now work with the full summary()/ datestamp()/tidy()/autoplot() pipeline, not just summary()/ tidy(). Found first in radf_tt_cv() (a user-reported datestamp() crash on radf_sign() prompted checking all three GLS-demeaned-family functions), then confirmed the identical fix applies to the other two. Validated per function: badf_cv‘s last row is bit-identical to adf_cv (a hard identity), empirical false-alarm rate at or below nominal (radf_tt 3.3%, radf_sign 5.5%, radf_sign_dm 3.5%, vs. 5% nominal), and detection power on an identical synthetic bubble in the same range as the established radf()/radf_mc_cv() baseline (16%): radf_tt 18%, radf_sign 20%, radf_sign_dm 8% (lower power a known, expected property of the sign-based tests’ heteroskedasticity invariance, not a validation concern) — see vignette("naming-and-analysis").

Performance

  • radf() is ~25x faster on typical sample sizes (exubercore v0.3.1): the recursive grid re-formed the full residual vector for every window, an O(n^3) total; it now keeps running cross-products and uses the closed-form SSR = y'y - b'X'y, making it O(n^2). n = 400 drops from ~140 ms to ~6 ms per path, so radf_mc_cv(), radf_wb_cv(), radf_sb_cv(), monitor(), dating_hlw(), radf_recovery() and every other loop over radf() speed up by the same factor. The regression is now parametrised as dy on (1, y_{t-1}, dy lags) (same t-statistic, on beta - 1), which is also more accurate than before at large n with large levels: ~3e-11 against lm() at n = 2000 where the old recursion was ~2e-8.
  • Parallel runs (options(exuber.parallel = TRUE)) reuse one worker cluster per session instead of starting and stopping a fresh future::multisession on every call – a few seconds of start-up overhead that used to dominate every small radf_mc_cv()/radf_wb_cv() job. The cluster is sized by exuber.ncores and stopped when the namespace unloads. exuber.parallel now defaults to interactive(): scripts, knitr and R CMD check run serially unless they opt in, so a batch job no longer pays worker start-up (nor leaves worker connections open) for a handful of replications.
  • radf()’s panel statistic used apply(bsadf, 1, mean), whose per-call dispatch overhead scales with the number of rows rather than staying fixed – 65x slower than the equivalent rowMeans(bsadf) at n = 100, 289x slower at n = 1000, and the single largest piece of radf()’s runtime at every sample size tested. Replaced with rowMeans(bsadf) (identical result, tests/testthat unchanged, 879 passing).

API consistency

  • One significance-level convention across the whole package: every function that took a level argument now takes sig_lvl on the 0-100 scale already used by datestamp()/autoplot() (sig_lvl = 95 = a 5% test / 95% confidence). Affected (all unreleased, so no shims): lbi_test(), monitor_lbi() (0.95 -> 95, 0.975 -> 97.5, …), ssu_test(), monitor(), monitor_cusum(), rootstamp() (was a 0.95-style confidence level), quantile_test()/monitor_quantile() (already 0-100, renamed only), and cobubble_test() (was a size, level = 0.05; now sig_lvl = 95). A shared assert_sig_lvl() makes sig_lvl = 0.95 an immediate error everywhere rather than a silently wrong quantile.
  • monitor(adflag = ) -> lag, matching radf(); monitor_cusum(N = ) -> h, matching every other kernel-bandwidth argument; cobubble_test(lags = ) -> lag_grid, so lag/lags can no longer be confused (mirrors quantile_test()’s tau/tau_grid).
  • cobubble_test() and radf_sbz_union() now return cobubble_test_obj/ radf_sbz_union_obj, the _obj suffix every other standalone class already carried.
  • dating_pdc() and radf_sbz() gained their own print() methods (the former fell through to print.data.frame, hiding its attributes; the latter printed as a plain radf).
  • scale_exuber_manual(size_values = ) is deprecated in favor of linewidth_values (ggplot2 >= 3.4.0’s linewidth aesthetic replaces size for lines; the deprecation warning ggplot2 emitted from every autoplot() is gone). autoplot(include_negative = ), deprecated since 1.0.0, is now actually forwarded to nonrejected instead of ignored.
  • radf_tt()/radf_tt_cv()/monitor_quantile() carry the same experimental badge as the other new methods; the “does not plug into autoplot” note on every standalone function was wrong (each has had its own autoplot() method) and now says so.

Other

  • rootstamp() — confidence interval and doubling time on the explosive root, via S3 dispatch: the default method fits a single sub-sample, the radf_obj method runs every datestamp() episode at once (previously three separate functions – explosive_root(), root_ci(), root_ci_datestamp() – consolidated before release).

  • Documentation: every autoplot() and augment() method now has its own help page instead of sharing one with the function it plots/tidies (?autoplot.monitor_cusum_obj, ?augment.radf_obj, …). The pkgdown reference index is reorganized into per-function subsections so each function is listed next to the methods that consume its output.

  • sim_vol_break() – i.i.d. Gaussian innovations whose standard deviation shifts permanently at a chosen break fraction, the non-stationary volatility DGP (Cavaliere & Taylor 2007) the volatility-robust tests are actually built for, unlike stationary GARCH. sim_ps1() gained the same e innovation-injection argument sim_psy1() already had, and the c/c1/c2 arguments of sim_psy1()/sim_ps1() now accept any positive scalar (as documented), so a fixed explosive root is expressible directly (c = 0.04, alpha = 0).

  • Examples and vignettes now demonstrate each method on the DGP it targets rather than on sim_data throughout: volatility-robust tests (radf_tt(), radf_kp(), radf_sign(), radf_sbz(), radf_wb_cv(), monitor_cusum(type = "kernel"), dating_pdc(type = "wls")) on a volatility break, cobubble_test()/contagion_reg() on sim_coexplosive(), radf_common() on sim_common(), ssu_test() on a stochastic root, quantile_test()/monitor_quantile() on heavy-tailed innovations, dating_*()/radf_recovery() on sim_ps1(), and every monitor on a bubble that starts after its training window. Hand-rolled cumsum(rnorm()) DGPs in the vignettes are replaced by the package’s own generators.

Bug fixes

  • radf_sb_cv()/radf_sb_distr() panel critical values were wrong in every release since 0.1.0. The bootstrap loop overwrote the per-series BSADF path instead of summing it, so the panel null distribution was the last series’ BSADF divided by the number of series rather than the cross-sectional mean. The error is invisible for ~5-series panels (the two quantities happen to coincide) but the test was oversized for narrow panels (8% at nominal 5% for 2 series) and had no power for wide ones (0% rejection under H0 and under the alternative for 10 series). Fixed; empirical size is now 6% / 5% / 2.5% for 2 / 5 / 10 series. A regression test pins the identity that a panel of identical copies has the same bootstrap distribution as the single series.

  • datestamp()/autoplot()/autoplot2()’s sig_lvl argument now actually controls whether a series counts as rejecting the null. Previously, diagnostics.radf_obj() (used internally to decide which series get dated/plotted at all) hard-coded the 95% critical value for that decision regardless of sig_lvl, so e.g. datestamp(x, cv, sig_lvl = 90) could throw "Cannot reject H0 at the 5% significance level" for a series that clearly rejects at the 10% level the caller asked for – sig_lvl only ever reached the within-series episode threshold curve, never the series-eligibility gate. diagnostics() gained a sig_lvl argument (default 95, so default-call behavior is unchanged) and datestamp()/autoplot()/autoplot2() now thread their own sig_lvl through to it.

  • radf() (and everything built on it) errored with subscript out of bounds on a named numeric vector, e.g. a prcomp() score column – which is exactly what radf_common() feeds it – because the names leaked into the internal NA-edge bookkeeping.

  • The default-cv range check said the store covers n <= 5000; it covers n <= 4000 and lag <= 4, and now says so before trying the network.

  • sim_psy1()’s seed argument now also covers a generator passed lazily to e/coef_noise (e.g. sim_psy1(n, seed = 1, e = sim_vol_break(n - 1))); previously those were forced before the seed was set.

exuber 1.1.0

CRAN release: 2025-08-31

  • Fixed targets not in the package itself nor in the base packages to use package anchors, i.e., use

exuber 1.0.1

CRAN release: 2023-02-12

Maintenance release to accommodate breaking changes in dplyr 1.1.0.

exuber 1.0.0

CRAN release: 2022-08-19

This first major release accompanies the publication of an article in the Journal of Statistical Software:

Vasilopoulos, K., Pavlidis, E., & Martínez-García, E. (2022). exuber: Recursive Right-Tailed Unit Root Testing with R. Journal of Statistical Software, 103(1), 1–26. https://doi.org/10.18637/jss.v103.i10

augment method for radf_obj and radf_cv

  • New arg trunc

  • Fixed inconsistencies among functions.

  • Now radf stores the data that are later can be accessed with mat+

  • Advanced features on datestamping: New columns that indicate:

    • Signal
    • Peak
    • Ongoing
    • Nonrejected

  • New datestamping procedure rev_radf etc.

  • New bootstrap procedure radf_wb_cv2 and radf_wb_distr2

  • New coloring convention for plotting ds and obj classes

Bug Fixes

  • Now autoplot can include periods that have an ongoing bubble

exuber 0.4.2

CRAN release: 2020-12-18

  • Include printing methods for radf_obj and radf_cv.
  • Removed unused class definitions.
  • Using progress package for progress_bar.

exuber 0.4.1

CRAN release: 2020-05-12

Maintenance release for compatibility with dplyr v1.0.0.

exuber 0.4.0

CRAN release: 2020-05-04

Design

We have the following design in mind for future scalability. If you want make inference about radf models, then the estimation can be achieved with radf() function and return an object of class radf_obj, and the critical values can be achieved with radf_*_cv() and return an object of class radf_cv.

Breaking changes

exuberdata

  • We created a new package called exuberdata that accommodates critical values for up to 2000 observations. Critical values can be examined with exuberdata::radf_crit2. The package is created through drat R archive Template, and can be easily installed with install.packages('exuberdata', repos = 'https://kvasilopoulos.github.io/drat/', type = 'source') or through install_exuberdata wrapper function that is provided in exuber.

Improvements

  • The package zoo has been used as a dependency to import the method index(). We made the decision to remove zoo and create a new method index() internally.

exuber 0.3.0

CRAN release: 2019-07-15

Breaking changes

  • Changed opt_bsadf = conservative for the simulated critical values (crit), also reduced the size of the crit from 700 to 600 due to package size restrictions.
  • sim_dgp1() and sim_dgp2() have been renamed to sim_psy1() and sim_psy2() to better describe the origination of the dgp.
  • sim_dgp1() and sim_dgp2() have been soft-deprecated.
  • autoplot_radf() arranges automatically multiple graphs, to return to previous behavior we included the optional argument arrange which is set to TRUE by default.

Three new functions have been added to simulate empirical distributions for:

  • mc_dist(): Monte Carlo
  • wb_dist(): Wild Bootstrap
  • sb_dist(): Sieve Bootstrap

and a function that can calculate the p-values calc_pvalue() given the above distributions as argument.

Also methods tidy() and autoplot() have been added to turn the object into a tidy tibble and draw a particular plot with ggplot2, respectively.

New features

  • tidy() methods for objects of class radf, cv.
  • augment() methods for objects of class radf and cv.
  • augment_join() to combine object radf and cv into a single data.frame.
  • glance() method for objects of class radf.

Improvements

  • New printing output for the functions summary(), diagnostics() and datestamp().
  • New improved progressbar with more succinct printing for wb_cv()
  • seed argument to functions that are using rng. Also the option to declare a global seed for reproducibility with the option(exuber.global_seed = ###)

Bug Fixes

  • sb_cv() and wb_cv()now can parse data that contain a date-column. Similarly, to what radf() is doing.

exuber 0.2.1.9000

  • Website development

exuber 0.2.1

CRAN release: 2019-03-01

  • Changed DESCRIPTION to include sb_cv reference.
  • Renamed boolean to dummy from datestamp and diagnostics.
  • datestamp dummy is now an attribute.

exuber 0.2.0

CRAN release: 2019-02-04

Options

Some of the arguments in the functions were included as options, you can set the package options with e.g. options(exuber.show_progress = TRUE).

  • parallel option boolean, allows for parallel in critical values computation.
  • ncores option numeric, sets the number of cores, defaults to max - 1.
  • show_progress option boolean, allows you to disable the progress bar, defaults to TRUE.

New features

  • Panel estimation in radf()
  • Added sb_cv() function: Panel Sieve Bootstrapped critical values
  • Default critical values are supplied directly into summary(), diagnostics, datestamp() and autoplot(), without having to specify argument cv. The critical values have been simulated from mc_cv() function and stored as data. Custom critical values should be provided by the user with the option cv.
  • Added ggarrange() function, that can arrange a list of ggplot objects into a single grob.
  • Added fortify to arrange a data.frame from radf() function.

Improvements

  • Parallel and ncores arguments are now set as options.
  • Ability to remove progressbar from package options.
  • radf() can parse date from ts objects.
  • report() has been renamed into summary().
  • plot() has been renamed into autoplot().
  • plot() and report() are soft deprecated.

Bug Fixes

  • Progressbar appears in the beginning of the iteration
  • Plotting date now works without having to to include any additional plotting option