Description Usage Arguments Details Value Author(s) See Also Examples
Calculate and plot achieved precision versus recall using adjusted p-values (padj) and known true positives (labels).
1 2 3 4 5 6 7 8 | precREC(object, padj, ...)
## S4 method for signature 'SimResults,missing'
precREC(object, threshold=c(0.99, 0.95, 0.9), transformation="1-x", plot=TRUE, ...)
## S4 method for signature 'missing,ANY'
precREC(padj, labels, threshold=c(0.99, 0.95, 0.9), transformation="1-x", plot=TRUE, ...)
## S4 method for signature 'SimResultsList,missing'
precREC(object, threshold=c(0.99, 0.95, 0.9), transformation="1-x", plot=TRUE, ...)
|
object |
An object of |
padj |
A vector or matrix containing the adjusted p-value. |
labels |
A numeric vector indicating class labels (positives with 1, negatives with 0). For more details, see |
threshold |
Numeric value of the adjusted p-value threshold (cutoff), or vector of multiple thresholds, used to determine the significance of tests. |
transformation |
The transformation of |
plot |
Logical, should precision-recall values be plotted. |
... |
Optional arguments for plotting (see below for more details).
|
The function precREC
calculates the achieved precision (positive predictive value (PPV) or 1-FDR) and recall (true positive rate (TPR)) given a particular threshold, control precision. Several thresholds can be assessed simultaneously (e.g., threshold=c(0.99,0.98,0.97,0.96,0.95)
).
Most of the graphical parameters such as col
, cex
, pch
and etc from par
, can be directly passed to plot
for precision-recall values. point.type
, letter indicating how precision-recall values should be plotted: "b" for both points and lines; "p" for points only; and "l" for lines only. col.line
and lwd.line
, col
and lwd
of line connecting precision-recall points, if point.type
is either "b" or "l". lwd.threshold
, lty.threshold
and col.threshold
are lwd
, lty
and col
referred to the lines drawn for 'threshold'.
The precision is considered to be controlled if the precision (1-FDR) falls above the threshold it is assessed at, and precision-recall points are filled-in. precision-recall points are unfilled, or empty, if the precision is not controlled and is equal to or less than the cutoff it is assessed at. Note that the fill-unfill option is only compatible with pch
values of 21 to 25. Visually, this means that for a given threshold, a point (of the same color) will be filled if it falls left of the threshold line, and will be unfilled if it sits to the right of the threshold line.
To increase the flexibility of plots, you can use argument add
. If add=TRUE
, plots are added to current one combining different rocX
curves together into one figure.
legend
is a list including all the arguments from function legend
(see help(legend)). If it is NULL, do not add legends to plots.
If input is an SimResultsList
object, the average performance would be plotted (see example 3).
Invisibly, an S4 object of class precREC-class
containing the threshold value(s) assessed, and the associated precision, achieved precision, PREC, and true positive rates, TPR.
Xiaobei Zhou and Mark D. Robinson
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | # example1
data(Pickrell)
re <- SimResults(pval=Pickrell$pval, labels=Pickrell$labels)
p <- precREC(re, plot=TRUE)
p <- precREC(re, plot=TRUE, legend=NULL)
plot(p, cex=2, pch=c(23,25), col=1:3, main="precREC plot",
lwd.line=2, cex.axis=1.5, col.threshold=4, lwd.threshold=2)
# example2
data(calibration)
padj <- calibration$padj
labels <- calibration$labels
precREC(padj=padj, labels=labels, xlim=c(0,.3))
# example3
data(PickrellList)
rel <- SimResultsList()
for(i in 1:5)
rel[[i]] <- SimResults(pval=PickrellList[[i]]$pval, labels=PickrellList[[i]]$labels)
powerFDR(rel)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.