NEWS

---

cointsmall 1.0.4 (2026-09-28)

• Bug fix: the critical values did not come from Trinh (2022). The model "o" table was shifted by one regressor (the m = 1 row held the Dickey-Fuller values of a single series), and the tables for one and two breaks were not the published response surfaces. The 5% critical values are now computed from the response surfaces in Table 13 of Trinh (2022), for m = 1, 2, 3 regressors; they reproduce Table 1 of the paper.
• Bug fix: the p-value was extrapolated from these tables and could exceed 1 (16.29 in one test case). Trinh (2022) publishes the 5% quantile only, so pvalue, cv01 and cv10 are now NA, level must be 5, and cointsmall_cv() stops for other levels or more than three regressors.
• The author of the methodology is Jerome Trinh; the reference is corrected, and the contributor entry in Authors@R, which named another person and did not correspond to any contributed code, is removed.
• The ADF* statistic and break dates are unchanged; they agree with the Stata command cointsmall (SSC) on the same data.

cointsmall 1.0.3

• Removed a DOI wrongly attached to the MacKinnon (2010) working paper; the citation text is unchanged. No changes to code.

cointsmall 1.0.0

• Initial CRAN release.

Features

• cointsmall(): Main function for cointegration tests with structural breaks in small samples
• cointsmall_combined(): Combined testing procedure evaluating all model specifications
• cointsmall_cv(): Function to retrieve critical values for different model configurations

Model Specifications

• Model "o": No structural break (standard Engle-Granger test)
• Model "c": Break in constant only (Gregory-Hansen style level shift)
• Model "cs": Break in constant and slope (regime change model)

Supported Options

• 0, 1, or 2 structural breaks
• Break date selection via minimum ADF statistic or minimum SSR
• Adjustable trimming parameter for break date search
• Automatic lag selection for ADF test using BIC
• Small-sample adjusted critical values via response surface methodology

References

• Based on Trinh (2022) "Testing for cointegration with structural changes in very small sample"