alphaByUseFunction {GroupSeq} | R Documentation |
This function computes type I error to spend at each analysis using an alpha-spending function.
alphaByUseFunction(whatSpendingFunctionIsUsed, n, alpha, phi, OneOrTwoSidedBounds, t)
whatSpendingFunctionIsUsed |
indicates type I error spending rate function and is an integer value. |
n |
is the number of interim analyses so far |
alpha |
is the desired overall size |
phi |
optional Parameter referring to Power Family |
OneOrTwoSidedBounds |
One-(OneOrTwoSidedBounds==1) or Two-sided bounds(OneOrTwoSidedBounds==2) |
t |
is the vector of information times |
The use function, or type I error spending rate, corresponds to the
Alpha 1 star, Alpha 2 star and Alpha 3 star in Lan & DeMets(1983).
probExit |
the vector of type I error (e.g. exit probabilities) spent at each analysis. |
This function is meant to be called in context of using groupseq
The parameter whatSpendingFunctionIsUsed is an integer value. At the moment there are
5 functions available - so only parameter from 1-5 will work:
Approximated by spending function:
(1) O'Brien-Fleming Type
(2) Pocock Type
(3) Power family: $α t^{phi}$
(4) Hwang-Shih-DeCani fammily
Classic Design:
(5) Pocock - the real Pocock Bounds
Roman Pahl
Lan & DeMets 1983 Biometrika http://www.biostat.wisc.edu/landemets/
groupseq
, asHSdeCani
, asPowerFamily
asOBF
, asPocock
## One-sided using O'Brien-Fleming Type alphaByUseFunction(1,3,0.05,0,1,c(0.33,0.66,1.0)) ## Two-sided using Power family: $\alpha t^{\phi}$ with $\phi$==1 alphaByUseFunction(3,3,0.05,1,2,c(0.33,0.66,1.0))