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
Minor improvements and
fixes
Fix inaccuracies about theoretical properties of the variance
estimator in package vignette
dfadjust 1.0.4
Minor improvements and
fixes
Adjust tolerance in unit tests so there are no issues on M1 Mac
dfadjust 1.0.3
Minor improvements and
fixes
Fix incorrect computation of p-values in the
print.dfadjustSE method
dfadjust 1.0.2
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
Minor improvements and
fixes
Fix problem with failing tests when platform didn’t use long
double
dfadjust 1.0.0
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)