COBRAData: 'COBRAData' object and constructor

In csoneson/iCOBRA: Comparison and Visualization of Ranking and Assignment Methods

COBRAData object and constructor

Description

The COBRAData class contains slots to hold calculated p-values, adjusted p-values and general 'scores' for a set of features, as well as s-values (see Stephens (2017)). The slots can contain values from multiple methods, and each method can contribute to one or more slots. The class also contains a slot giving the 'truth' (a binary assignment and/or a continuous score) for each feature, as well as additional annotations that can be used to stratify the performance calculations.

Usage

COBRAData(
  pval = data.frame(),
  padj = data.frame(),
  score = data.frame(),
  sval = data.frame(),
  truth = data.frame(),
  object_to_extend = NULL
)

COBRAData_from_text(truth_file, result_files, feature_id)

COBRAData_to_text(cobradata, truth_file, result_files, feature_id)

Arguments

pval	A data frame with features as rows and methods as columns, containing nominal p-values. Missing values (NAs) are allowed. The row names should be feature names.
padj	A data frame with features as rows and methods as columns, containing adjusted p-values. Missing values (NAs) are allowed. The row names should be feature names.
score	A data frame with features as rows and methods as columns, containing generic scores. In case of comparison to a binary truth, larger values of the scores should correspond to 'more significant' features. Missing values (NAs) are allowed. The row names should be feature names.
sval	A data frame with features as rows and methods as columns, containing s-values (analogous to q-values, but for sign errors, see Stephens (2017)). Missing values (NAs) are allowed. The row names should be feature names.
truth	A data frame with features as rows columns containing feature annotations such as, e.g., binary and continuous truths and additional annotations that can be used to stratify the performance calculations. The row names should be feature names.
object_to_extend	A COBRAData object to extend with the provided information.
truth_file	A character string giving the path to a file with true labels and other feature annotations.
result_files	A character vector giving path(s) to file(s) with results (p-values, adjusted p-values, s-values, scores) for one or more methods. The column names of these files must be of the form "method:measure", where measure is one of P, adjP, S or score, depending on what is given in the column.
feature_id	A character string giving the name of the column in the truth and result files that encodes the feature identifier.
cobradata	A COBRAData object

Details

If adjusted p-values are missing for some methods, for which nominal p-values are available, the adjusted p-values can be calculated using the calculate_adjp function.

The text files generated by COBRAData_to_text can be used as input to iCOBRAapp, when it is called without an input argument.

Value

COBRAData and COBRAData_from_text return a COBRAData object.

Author(s)

Charlotte Soneson

Examples

## Empty COBRAData object:
COBRAData()

## COBRAData object from individual data frames
set.seed(123)
pval <- data.frame(m1 = runif(100), m2 = runif(100),
                   row.names = paste0("F", 1:100))
truth <- data.frame(status = round(runif(100)),
                    row.names = paste0("F", 1:100))
cobradata <- COBRAData(pval = pval, truth = truth)

csoneson/iCOBRA documentation built on May 3, 2024, 5:23 a.m.