steppedwedge 1.1.0
Major changes
analyze() supports the Delayed Constant Treatment (DCT)
model.analyze() supports cluster-level random treatment
effects via re = "tx".analyze() can adjusts for covariates passed via
load_data(covariates = ).analyze() supports robust variance estimation via
params(var_est = "robust").load_data() accepts an exposure_time
variable (in addition to automatically calculating exposure time when
only a binary treatment variable is passed).plot_effect_curves() visualizes estimated treatment
effect curves.plot_clusters() visualizes cluster-level data or fitted
analysis objects.
Minor changes
- The
offset argument of analyze() has moved
to load_data(). Passing offset through
params() is deprecated.
load_data() now stores time_type as an
attribute rather than a dataframe column. Access it with
attr(data, "time_type").params(return_ncs=TRUE) returns the natural cubic
spline transformation matrix and the transformed coefficients and
covariance matrices.- Documentation and vignettes include examples for binary outcomes,
DCT models, and cluster plots.
- The binomial outcome dataset is now
sw_data_example_binom.
- For spline-based model components, the number of knots can be
configured via
params(), including the exposure time spline
of the NCS model (n_knots_exp) and the calendar time trend
spline (n_knots_cal).
- Fix to how
estimand_time is coded in the TEH
model.
- An informative error is thrown if the user tries to use a robust
variance estimator with a TEH model.
print() formats and displays estimates, standard
errors, and p-values for analysis objects.
steppedwedge 1.0.0