The adoptr package provides functionality to explore custom optimal two-stage designs for one- or two-arm superiority tests. More than two arms can be compared via chi-squared tests or ANOVA. For more details on the theoretical background see <doi:10.1002/sim.8291> and <doi:10.18637/jss.v098.i09>. adoptr makes heavy use of the S4 class system. A good place to start learning about it can be found here.
For a sample workflow and a quick demo of the capabilities, see here.
A more detailed description of the background and the usage of adoptr can be found here or here <doi:10.18637/jss.v098.i09> .
A variety of examples is presented in the validation report hosted here.
adoptr currently supports TwoStageDesign,
GroupSequentialDesign-class, and OneStageDesign-class.
The implemented data distributions are Normal, Binomial,
Student, Survival, ChiSquared (including
Pearson2xK and ZSquared) and ANOVA.
Both ContinuousPrior and PointMassPrior are
supported for the single parameter of a DataDistribution.
See Scores for information on the basic system of representing
scores.
Available scores are ConditionalPower,
ConditionalSampleSize, Power, and
ExpectedSampleSize..