Scores

Inadoptr scores are used to assess the performance of a design.This can be done either conditionally on the observed stage-one outcomeor unconditionally.Consequently, score objects are either of classConditionalScore orUnconditionalScore.

Usage

expected(s,data_distribution,prior,...)# S4 method for class 'ConditionalScore'expected(s,data_distribution,prior, label=NA_character_,...)evaluate(s,design,...)# S4 method for class 'IntegralScore,TwoStageDesign'evaluate(s,design, optimization=FALSE, subdivisions=10000L,...)

Arguments

s

Score object

data_distribution

DataDistribution object

prior

aPrior object

...

further optional arguments

label

object label (string)

design

object

optimization

logical, ifTRUE uses a relaxation to realparameters of the underlying design; used for smooth optimization.

subdivisions

maximal number of subdivisions when evaluating an integralscore using adaptive quadrature (optimization = FALSE)

Value

No return value. Generic description of classScore.

Details

All scores can be evaluated on a design using theevaluate method.Note thatevaluate requires a third argumentx1 forconditional scores (observed stage-one outcome).AnyConditionalScore can be converted to aUnconditionalScoreby forming its expected value usingexpected.The returned unconditional score is of classIntegralScore.

See also

Examples

design<-TwoStageDesign(  n1=25,  c1f=0,  c1e=2.5,  n2=50,  c2=1.96,  order=7L)prior<-PointMassPrior(.3,1)# conditionalcp<-ConditionalPower(Normal(),prior)expected(cp,Normal(),prior)#> E[Pr[x2>=c2(x1)|x1]]<Normal<two-armed>;PointMass<0.30>>evaluate(cp,design, x1=.5)#> [1] 0.3227581# unconditionalpower<-Power(Normal(),prior)evaluate(power,design)#> [1] 0.3269562evaluate(power,design, optimization=TRUE)# use non-adaptive quadrature#> [1] 0.3269562