as.cst: Create a data frame from a Collection of Supervised Trials...

In davidavdav/ROC: Compute structures to compute ROC and DET plots and metrics for 2-class classifiers.

Create a data frame from a Collection of Supervised Trials for ROC analysis

Description

These functions produce a cst object, which is a data frame blessed to be used for various ROC analysis tools. A cst object is an arbitrary data frame with at least two columns: one named target, a logical indicating to which class the trial belongs (in detection theory named a "target" vs "non-target" trial, but other fields have different names such as "same-source" vs "difference source"). The second required column is score, which is either an ordered factor or a numeric, where higer numbers indicate a higher likelihood that the trial is of the kind target==TRUE.

as.cst takes as input a data frame, checks for the columns and their types, and blesses it to be of class cst. cst.tnt takes two arguments: one for target scores and one for non-targets scores, and it produces a data frame of class cst

Usage

as.cst(df)
cst.tnt(tar, non)

Arguments

df	A data frame, containing columns target and score
tar, non	Target and non-target scores. Either a vector of values, or a file name that is scanned for a vector of values.

Details

The data frame df may contain additionmal columns, which are maintained in further processing. They can be used for conditioning the ROC analysis, e.g., in plot.cond.

The arguments tar and non can either refer to a vector of values or to a file name. In the latter case, the file is opened and an attempt is made to read numeric values from the file using scan.

Value

A data frame of class cst with at least the columns target and score of the right type.

Author(s)

David A. van Leeuwen

Examples

## create normally disributed target and nontarget scores around 0 and 1.
df <- data.frame(target=rep(c(F,T),each=100), score=rnorm(200, mean=rep(c(0,1), each=100)))
x <- as.cst(df)
roc(x)
## read in data from vectors
y <- cst.tnt(rnorm(100, 1), rnorm(1000, 0))
## read in data from files
z <- cst.tnt("targets.txt", "nontargets.txt")

davidavdav/ROC documentation built on Sept. 8, 2023, 2:39 p.m.