Scores: Scores
In adoptr: Adaptive Optimal Two-Stage Designs in R

Description Usage Arguments Details See Also Examples

In adoptr scores are used to assess the performance of a design. This can be done either conditionally on the observed stage-one outcome or unconditionally. Consequently, score objects are either of class ConditionalScore or UnconditionalScore.

expected(s, data_distribution, prior, ...)

## S4 method for signature 'ConditionalScore'
expected(s, data_distribution, prior, label = NA_character_, ...)

evaluate(s, design, ...)

## S4 method for signature 'IntegralScore,TwoStageDesign'
evaluate(s, design, optimization = FALSE, subdivisions = 10000L, ...)

`s`	`Score` object
`data_distribution`	`DataDistribution` object
`prior`	a `Prior` object
`...`	further optional arguments
`label`	object label (string)
`design`	object
`optimization`	logical, if `TRUE` uses a relaxation to real parameters of the underlying design; used for smooth optimization.
`subdivisions`	maximal number of subdivisions when evaluating an integral score using adaptive quadrature (optimization = FALSE)

All scores can be evaluated on a design using the evaluate method. Note that evaluate requires a third argument x1 for conditional scores (observed stage-one outcome). Any ConditionalScore can be converted to a UnconditionalScore by forming its expected value using expected. The returned unconditional score is of class IntegralScore.

ConditionalPower, ConditionalSampleSize, composite

design <- TwoStageDesign(
  n1    = 25,
  c1f   = 0,
  c1e   = 2.5,
  n2    = 50,
  c2    = 1.96,
  order = 7L
)
prior <- PointMassPrior(.3, 1)

# conditional
cp <- ConditionalPower(Normal(), prior)
expected(cp, Normal(), prior)
evaluate(cp, design, x1 = .5)

# unconditional
power <- Power(Normal(), prior)
evaluate(power, design)
evaluate(power, design, optimization = TRUE) # use non-adaptive quadrature