Estimation | Analysis | Plotting

Monitoring and testing | One bubble, five monitors | Which to reach for

What "experimental" means here | radf_recovery(): dating a collapse and a recovery | datestamp(option = "svadf"): a preprint | Using them responsibly

What this page covers | The naming scheme | What actually plugs into summary()/datestamp()/tidy()/autoplot() | Full support: radf_common(), radf_kp(), radf_tt(), radf_sign(), radf_sign_dm(), radf_sbz() | Own print() and autoplot(): everything else | Summary

One autoplot() per object | The radf() plot | Changing the appearance | Building your own plot | Distributions | Which to reach for

Why another test | Basic usage | What is estimated | Dating and plotting a detected bubble

The shared problem | Sign-based: radf_sign() | Kernel-purged: radf_kp() | WLS + kernel volatility: radf_sbz() | Union-of-rejections: radf_sbz_union() | Which to reach for

Why not just use radf() | A fixed root, which lbi_test() detects | A varying root, which ssu_test() detects and lbi_test() misses | Testing a quantile instead of the mean | Which to reach for

Two different questions about two series | cobubble_test(): are these the same bubble? | contagion_reg(): how strongly do they co-move, and when? | Which to reach for

Why these exist alongside datestamp() | A single bubble, four estimates | Which to reach for

A different question from "is there a bubble" | Detect, date, then estimate the root | Every episode at once

What the sim_*() family is for | The PSY experiment | Plotting a simulated series | Swapping the innovations | Multi-series processes | Which to reach for

The map | Step 1: baseline fit and three diagnostics | Step 2: serially correlated errors → ivx_ar() | Step 3: heteroskedasticity → robust = TRUE | Step 4: finite-sample size → ivx_boot() and ivx_ra() | Step 5: power → lag_y = TRUE | Step 6: what is the null? | Step 7: cross-check with a non-IVX test | The worked example, in one table | Summary | References

The problem | The IVX instrument | Tuning | Long horizons | Lag augmentation for power | Heteroskedasticity-robust standard errors | Replication of Kostakis et al. (2015) | Caveats | References

Setting | The test | Tuning | Block bootstrap under conditional heteroskedasticity | Caveats | References

Bonferroni Q-test | Augmented regression method | Idea | Check | Caveats | Control-function regression | Hybrid t-test | Unified empirical likelihood test | References

Idea | Procedure | Replication: size in the paper's Monte Carlo | Long horizons: the transformed regression | Caveats and implementation notes | References

Instruments | The statistic | Replication of Table 1 | Caveats | References

Subsample statistic | Interpretation and caveats | References

Model | Procedure | Caveats | References

Why bootstrap | The two schemes | Replication of DGRT (2023), Table 4 | Caveats | References

Model and statistic | Validation | Caveats | References

References