The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.
beck_banks dataset has been removed, together with
its licence note. The examples and the vignette now illustrate the
always-treated error and estimation without a never-treated group on
sim_panel.0 and period
0 lies inside the observed window,
gmm_staggered() now stops with an error. Previously
0 was read both as a treated cohort, with cohort-by-time
effects estimated for it, and as the never-treated control group. Code
never-treated units with any value after the last period instead.gmm_staggered() now stops with an error naming the
covariate and the cohort. Previously every estimate was returned as
NA, with a misleading warning that no weighted step could
be completed.gmm_j_test() is now explicitly a test of
unconditional parallel trends. Its placebo moments were never
covariate-adjusted; after a fit with covar, the printed
result now warns that it does not test the covariate-conditional
parallel trends assumption, summary() of such a fit says
the same, and the returned object records scope and
covar.First release.
gmm_staggered(), replacing
the earlier gmm_staggered_I(),
gmm_staggered_II() and gmm_staggered_III().
The covariance model is chosen with weighting, one of
"pooled_toeplitz", "cohort_toeplitz" or
"unrestricted" (GMM-T, GMM-HT and GMM-U in the paper).did package convention:
yname, tname, idname,
gname.never_treated = NULL detects a never-treated group
automatically, replacing has_nt = FALSE.staggered_gmm with
print, summary, coef,
vcov and confint methods.gmm_j_test(): the serial-correlation robust
over-identification test of parallel trends and no anticipation, in
full-set and local-pre-window forms.CW
(treated-observation weighting, as used throughout the paper) and
EW (cohort-equal weighting).NA with identified = FALSE, never as an
estimated zero, and are excluded from the parameter vector so that
partial identification no longer disables the efficient weighting for
every other cell.max_iter
now defaults to 100: the iteration can converge linearly at a slow rate,
and sim_panel under
weighting = "cohort_toeplitz" needs 86 steps.weighting = "unrestricted", because the returned estimates
are then the identity-weighted seed rather than a GMM estimate under the
requested weighting. The warning reports how many independent moment
directions the estimated weight supported, against the number of effects
to identify. Merely failing to converge after a weighted step did run
continues not to warn under "unrestricted", where it is
routine.beck_banks documentation now records its CC BY 4.0
licence and the thirteen always-treated states it contains.These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.