Reference index

Function names follow what the function computes, not the paper it came from. The prefix radf_ is reserved for functions built on the recursive ADF core. They either call radf() directly or reuse its badf and bsadf output (radf_tt(), radf_sign(), and the _cv and _mc critical-value functions). All other functions are named for what they do. A _test suffix marks a hypothesis test with its own null distribution (lbi_test(), ssu_test(), quantile_test(), cobubble_test()). A dating_ prefix marks a point-estimation or model-selection procedure with no formal test (dating_hls(), dating_hlw(), dating_knp(), dating_pdc()). A monitor_ prefix marks real-time or sequential detection (monitor_cusum(), monitor_lbi(), monitor_quantile() and monitor_radf()). We list monitor_radf() with the other monitors even though it reuses the badf and bsadf output of radf(), because a reader looking for that prefix wants the functions that monitor in real time. A root_ prefix marks confidence-interval inference on the magnitude of the explosive autoregressive root, which is a different question from whether it is present (rootstamp() has two S3 methods: the default method handles a single sub-sample, and the radf_obj method runs every datestamp() episode at once). Finally, contagion_reg() stands on its own as a point-estimation tool that performs no formal inference. The names are a guide and not a contract. For programmatic use, call exuber_functions(family = ...) instead of parsing names.

The dating_ family above is separate from datestamp() in the Analysis group below. datestamp() applies the PSY threshold-crossing rule and is a generic that works on any radf_obj and radf_cv pair. The dating_ functions are unrelated SSR/BIC dating procedures that take raw data directly and use no critical value.

Only results of class radf_obj can be passed to the shared Analysis, Tidying and Plotting generics below (summary(), datestamp(), tidy(), autoplot()). Other results print in their own format, which those generics do not fit. radf_common() and radf_kp() return the output of radf() unchanged, so all four generics work on them. radf_sign(), radf_sign_dm() and radf_tt() also carry the class, but only summary() and tidy() work on them at present. Their _cv() functions compute only the three scalar critical values that summary() needs, not the time-varying boundary that datestamp() and autoplot() require. This is a known gap. For the full picture with worked examples, see vignette("naming-and-analysis", package = "exuber").