NEWS

---

dfadjust 1.1.0 (2024-12-18)

New Features

• Allow argument ell to be shorter than covariate dimension. In this case, ell specifies which subset of covariates to compute standard errors for.

Minor improvements and fixes

• Use collapse::fsum instead of tapply calls to improve speed
• Check that covariates are not collinear, drop the collinear ones

dfadjust 1.0.5 (2023-10-04)

Minor improvements and fixes

• Fix inaccuracies about theoretical properties of the variance estimator in package vignette

dfadjust 1.0.4 (2022-04-24)

Minor improvements and fixes

• Adjust tolerance in unit tests so there are no issues on M1 Mac

dfadjust 1.0.3 (2021-08-10)

Minor improvements and fixes

• Fix incorrect computation of p-values in the print.dfadjustSE method

dfadjust 1.0.2 (2021-02-24)

Minor improvements and fixes

• Fix incorrect computation of CR2 variance estimator and degrees of freedom adjustment if data not sorted by cluster

dfadjust 1.0.1 (2019-12-16)

Minor improvements and fixes

• Fix problem with failing tests when platform didn't use long double

dfadjust 1.0.0 (2019-08-23)

New Features

• The function dfadjustSE implements small-sample degrees of freedom adjustment discussed in Imbens and Kolesár (2016), using both heteroskedasticity-robust and clustered standard errors. For clustered standard errors, the package implements both the Imbens and Kolesár (2016) and the Bell and McCaffrey (2002, Survey Methodology) degrees of freedom adjustments.
• This implementation can handle models with fixed effects, as well as datasets with a large number of observations (for heteroskedasticity-robust standard errors) or datasets with large clusters (for clustered standard errors)