merge_p_values {ActivePathways}

Merge a list or matrix of p-values

Description

Merge a list or matrix of p-values

Usage

merge_p_values(
  scores,
  method = "Fisher",
  scores_direction = NULL,
  constraints_vector = NULL
)

Arguments

scores

Either a list/vector of p-values or a matrix where each column is a test.

method

Method to merge p-values. See 'methods' section below.

scores_direction

Either a vector of log2 transformed fold-change values or a matrix where each column is a test. Must contain the same dimensions as the scores parameter. Datasets without directional information should be set to 0.

constraints_vector

A numerical vector of +1 or -1 values corresponding to the user-defined directional relationship between the columns in scores_direction. Datasets without directional information should be set to 0.

Value

If scores is a vector or list, returns a number. If scores is a matrix, returns a named list of p-values merged by row.

Methods

Eight methods are available to merge a list of p-values:

Fisher

Fisher's method (default) assumes that p-values are uniformly distributed and performs a chi-squared test on the statistic sum(-2 log(p)). This method is most appropriate when the columns in scores are independent.

Fisher_directional

Fisher's method modification that allows for directional information to be incorporated with the scores_direction and constraints_vector parameters.

Brown

Brown's method extends Fisher's method by accounting for the covariance in the columns of scores. It is more appropriate when the tests of significance used to create the columns in scores are not necessarily independent. Note that the "Brown" method cannot be used with a single list of p-values. However, in this case Brown's method is identical to Fisher's method and should be used instead.

DPM

DPM extends Brown's method by incorporating directional information using the scores_direction and constraints_vector parameters.

Stouffer

Stouffer's method assumes p-values are uniformly distributed and transforms p-values into a Z-score using the cumulative distribution function of a standard normal distribution. This method is appropriate when the columns in scores are independent.

Stouffer_directional

Stouffer's method modification that allows for directional information to be incorporated with the scores_direction and constraints_vector parameters.

Strube

Strube's method extends Stouffer's method by accounting for the covariance in the columns of scores.

Strube_directional

Strube's method modification that allows for directional information to be incorporated with the scores_direction and constraints_vector parameters.

Examples

  merge_p_values(c(0.05, 0.09, 0.01))
  merge_p_values(list(a=0.01, b=1, c=0.0015, d=0.025), method='Fisher')
  merge_p_values(matrix(data=c(0.03, 0.061, 0.48, 0.052), nrow = 2), method='Brown')