It is often challenging to strongly control the family-wise type-1 error rate in the group-sequential trials with multiple endpoints (hypotheses). The inflation of type-1 error rate comes from two sources (1) repeated testing individual hypothesis and (2) simultaneous testing multiple hypotheses. The MultiGroupSequential package is intended to help researchers to tackle this challenge.
The procedures provided include the sequential procedures described in Luo and Quan (2023) and the graphical procedure proposed by Maurer and Bretz (2013). Luo and Quan (2023) describes three procedures and functions to implement these procedures
seqgspgx()
implements a sequential graphical procedure
based on the group-sequential p-values.seqgsphh()
implements a sequential Hochberg/Hommel
procedure based on the group-sequential p-values.seqqvalhh()
implements a sequential Hochberg/Hommel
procedure based on the q-values.In addition, seqmbgx()
implements the sequential
graphical procedure described in Maurer and Bretz (2013).
You can install the released version of MultiGroupSequential from CRAN:
install.packages("MultiGroupSequential")
For example, to use the sequential graphical procedure based on group sequential p-values.
pm
has
alpha
is the overall family-wise type-1 error
rate.W
is the weights of the graph assigned to each
hypothesis, whereas G
holds the transition matrix of the
graph.The procedures implemented here will usually give output list with elements:
rejected
: the index set of rejected hypothesesdecisionsm
: rejection decision for each hypothesis
(row) at each time point (column)cumdecisionsm
: cumulative rejection decision for each
hypothesis (row) at each time point (column)library(MultiGroupSequential)
seqgspgx(
pm = rbind(c(0.02, 0.03, 0.01), c(0.03, 0.04, 0.01)),
alpha = 0.025,
W = c(0.6, 0.4),
G = rbind(c(0, 1), c(1, 0))
)#> $rejected
#> [1] 1 2
#>
#> $decisionsm
#> [,1] [,2] [,3]
#> [1,] 0 0 1
#> [2,] 0 0 1
#>
#> $cumdecisionsm
#> [,1] [,2] [,3]
#> [1,] 0 0 1
#> [2,] 0 0 1