Multiple testing using graphs

Implements the graphical test procedure described in Bretz et al. (2009). Note that the gMCP function in the gMCP package performs the same task.

Usage

graphTest(
  pvalues,
  weights = NULL,
  alpha = 0.05,
  G = NULL,
  cr = NULL,
  graph = NULL,
  verbose = FALSE,
  test,
  upscale = FALSE
)

Arguments

pvalues

Either a vector or a matrix containing the local p-values for the hypotheses in the rows.

weights

Initial weight levels for the test procedure, in case of multiple graphs this needs to be a matrix.

alpha

Overall alpha level of the procedure. For entangled graphs alpha should be a numeric vector of length equal to the number of graphs, each element specifying the partial alpha for the respective graph. The overall alpha level equals sum(alpha).

G

For simple graphs G should be a numeric matrix determining the graph underlying the test procedure. Note that the diagonal need to contain only 0s, while the rows need to sum to 1. For entangled graphs it needs to be a list containing the different graph matrices as elements.

cr

Correlation matrix that should be used for the parametric test. If cr==NULL the Bonferroni based test procedure is used.

graph

As an alternative to the specification via weights and G one can also hand over a graphMCP object to the code. graphMCP objects can be created for example with the graphGUI function.

verbose

If verbose is TRUE, additional information about the graphical rejection procedure is displayed.

test

In the parametric case there is more than one way to handle subgraphs with less than the full alpha. If the parameter test is missing, the tests are performed as described by Bretz et al. (2011), i.e. tests of intersection null hypotheses always exhaust the full alpha level even if the sum of weights is strictly smaller than one. If test="simple-parametric" the tests are performed as defined in Equation (3) of Bretz et al. (2011).

upscale

Logical. If upscale=FALSE then for each intersection of hypotheses (i.e. each subgraph) a weighted test is performed at the possibly reduced level alpha of sum(w)*alpha, where sum(w) is the sum of all node weights in this subset. If upscale=TRUE all weights are upscaled, so that sum(w)=1.

Value

A vector or a matrix containing the test results for the hypotheses under consideration. Significant tests are denoted by a 1, non-significant results by a 0.

References

Bretz, F., Maurer, W., Brannath, W. and Posch, M. (2009) A graphical approach to sequentially rejective multiple test procedures. Statistics in Medicine, 28, 586--604

Bretz, F., Maurer, W. and Hommel, G. (2010) Test and power considerations for multiple endpoint analyses using sequentially rejective graphical procedures, to appear in Statistics in Medicine