getPseudovalue {BuyseTest}

Extract the pseudovalues of the Estimator

Description

Extract the pseudovalues of the estimator. The average of the pseudovalues is the estimate and their standard deviation the standard error of the estimate times a factor n (i.e. a t-test on their mean will give asymptotically valid confidence intervals and p-values).

Usage

getPseudovalue(object, statistic = NULL, endpoint = NULL)

## S4 method for signature 'S4BuyseTest'
getPseudovalue(object, statistic = NULL, endpoint = NULL)

Arguments

object

an R object of class S4BuyseTest, i.e., output of BuyseTest

statistic

[character] the type of statistic relative to which the pseudovalues should be computed. Can be "netBenefit", "winRatio", "favorable", or "unfavorable".

endpoint

[character] for which endpoint(s) the pseudovalues should be output? If NULL returns the sum of the H-decomposition over all endpoints.

Author(s)

Brice Ozenne

See Also

BuyseTest for performing a generalized pairwise comparison.
S4BuyseTest-summary for a more detailed presentation of the S4BuyseTest object.

Examples

set.seed(10)
n <- 250
d <- simBuyseTest(n)

e.BT <- BuyseTest(treatment ~ tte(eventtime,status,2) + bin(toxicity),
                 data = d, trace = 0)

#### net Benefit
pseudo <- getPseudovalue(e.BT)
summary(lm(pseudo~1))$coef
## asymptotically equivalent to
confint(e.BT, transformation = TRUE)
## (small differences: small sample corrections)

summary(lm(getPseudovalue(e.BT, endpoint = 1)~1))$coef

#### win Ratio
pseudo <- getPseudovalue(e.BT, statistic = "winRatio")
summary(lm(pseudo~1))$coef ## wrong p-value (should compare to 1 instead of 0)
## asymptotically equivalent to
confint(e.BT, statistic = "winRatio", transformation = TRUE)

#### favorable
pseudo <- getPseudovalue(e.BT, statistic = "favorable")
summary(lm(pseudo~1))$coef ## wrong p-value (should compare to 1/2 instead of 0)
## asymptotically equivalent to
confint(e.BT, statistic = "favorable", transformation = TRUE)

#### unfavorable
pseudo <- getPseudovalue(e.BT, statistic = "unfavorable")
summary(lm(pseudo~1))$coef ## wrong p-value (should compare to 1/2 instead of 0)
## asymptotically equivalent to
confint(e.BT, statistic = "unfavorable", transformation = TRUE)