allpairsTest: Tests for the Interaction Information for several pairs of...
In teisseyrep/interactionInfo: Statistical tests for Interaction Information

Description Usage Arguments Details Value Author(s) References See Also Examples

The function performs tests for the positeveness of Interaction Information (chi squared test, permutation test or hybrid test) for all pairs in the data set.

1	allpairsTest(x,y,calculateTest=TRUE,type="hybridTest",nbins=NULL,alpha=0.05,alpha0=0.05,B=1000,method="emp",trace=TRUE)

`x`	Matrix containing variables in the columns.
`y`	Additional variable.
`calculateTest`	Logical variable indicating whether the tests should be performed. If the value is FALSE, only the Interaction Information is computed. Default is TRUE.
`type`	Type of the test for positeveness of the Interaction Information. The possible values are: "hybridTest", "chisqTest" and "permTest". Default is "hybridTest".
`nbins`	Number of bins to be used for the discretization. By default the number of bins is set to (N)^(1/3) where N is the number of samples.
`alpha0`	Significance level of the initial test (see `hybridTest` for details), the default value is 0.05.
`alpha`	Significance level, the default value is 0.05.
`B`	Number of permutations, the default value is 1000.
`method`	The method used to estimate entropy. See function interinformation in R package infotheo for details.
`trace`	Logical variable indicating whether to print the analysis progress.

The function computes the Interaction Information I(xi,xj;y), where xi and xj are variables in x (columns of matrix x), for all pairs (xi,xj). If the variables xi, xj, y are not factors, they are discretized using 'discretize' function from R package 'infotheo'. Discretization is needed to calculate Interaction Information. The Interaction Information is computed using function interinformation from R package infotheo.

Data frame containing 4 (or 3 if calculateTest=FALSE) columns: index of the first variable, index of the second variable, Interaction Information, p-value.

Pawel Teisseyre

Pawel Teisseyre, Jan Mielniczuk, Michal J. Dabrowski, Detection of hidden associations and interactions in biomedical data using Interaction Information, manuscript, 2017.

hybridTest, chisqTest, permTest

#Example (strong interaction between 1 and 2 variable in x and no interactions between the remaining variables):
x= matrix(0,nrow=100,ncol=4)
x[,1]=c(rep(0,25),rep(1,25),rep(1,25),rep(0,25))
x[,2]=c(rep(0,25),rep(1,25),rep(0,25),rep(1,25))
x[,3] = rnorm(100)
x[,4] = rnorm(100)
y=c(rep(1,50),rep(0,50))
allpairsTest1=allpairsTest(x,y,calculateTest=TRUE)
print(allpairsTest1)