optimal_design: Compute an optimal adaptive design

In MatheMax/OptReSample: Computing Optimal Adaptive Designs

Description

optimal_design returns an optimal two-stage adaptive design.

Usage

optimal_design(effect, alpha, pow, effect_null = 0, sd = 1,
  standardized = T, t_approx = F, lagrange = T)

Arguments

effect	The effect on that the power is computed
alpha	The maximal type I error rate
pow	The minimal power
effect_null	The effect under the null hypothesis. Default is 0.
sd	The standard deviation. Default is 1.
standardized	Logical. Is effect already standardized? Default is T.
t_approx	Logical. Should a t-approximation be used? Default is F.
lagrange	Logical. Should a Lagrangian procedure be used? Default is T.

Details

The optimality criterion is given as the expected sample size under the alternative hypothesis. If standardized=T, the values of effect_null and sd are set to their default, even if they were specified differently.

The use of a t-approximation (t_approx=T) or a Lagrangian approach (lagrange=T) yields more precise results, but makes the calculation slowlier.

Note that there is no Lagrangian implementation of the t-approximation. Hence, the value of lagrange will be ignored, if t_approx is set to T.

Value

An object of class design

MatheMax/OptReSample documentation built on May 5, 2019, 8:14 a.m.