xPredictROCR: Function to assess the prediction performance via ROC and...

In Pi: Leveraging Genetic Evidence to Prioritise Drug Targets at the Gene and Pathway Level

Description

xPredictROCR is supposed to assess the prediction performance via Receiver Operating Characteristic (ROC) and Precision-Recall (PR) analysis. It requires three inputs: 1) Gold Standard Positive (GSP) targets; 2) Gold Standard Negative (GSN) targets; 3) prediction containing predicted targets and predictive scores.

Usage

xPredictROCR(
prediction,
GSP,
GSN,
rescale = TRUE,
plot = c("none", "ROC", "PR"),
verbose = TRUE,
font.family = "sans",
signature = TRUE
)

Arguments

prediction	a data frame containing predictions along with predictive scores. It has two columns: 1st column for target, 2nd column for predictive scores (the higher the better). Alternatively, it can be an object of class "pNode" (or "sTarget" or "dTarget") from which a data frame is extracted
GSP	a vector containing Gold Standard Positives (GSP)
GSN	a vector containing Gold Standard Negatives (GSN)
rescale	logical to indicate whether to linearly rescale predictive scores for GSP/GSN targets to the range [0,1]. By default, it sets to TRUE
plot	the way to plot performance curve. It can be 'none' for no curve returned, 'ROC' for ROC curve, and 'PR' for PR curve.
verbose	logical to indicate whether the messages will be displayed in the screen. By default, it sets to TRUE for display
font.family	the font family for texts
signature	a logical to indicate whether the signature is assigned to the plot caption. By default, it sets TRUE showing which function is used to draw this graph

Value

If plot is 'none' (by default), an object of class "pPerf", a list with following components:

• PRS: a data frame with 3 columns ('Precision', 'Recall' and 'Specificity')

• AUROC: a scalar value for ROC AUC

• Fmax: a scalar value for maximum F-measure

• ROC_perf: a ROCR performance-class object for ROC curve

• PR_perf: a ROCR performance-class object for PR curve

• Pred_obj: a ROCR prediction-class object (potentially used for calculating other performance measures)

If plot is 'ROC' or 'PR', it will return a ggplot object after being appended with the same components as mentioned above. If no GSP and/or GSN is predicted, it will return NULL

Note

AUC: the area under ROC F-measure: the maximum of a harmonic mean between precision and recall along PR curve

See Also

xPredictROCR

Examples

RData.location <- "http://galahad.well.ox.ac.uk/bigdata"
## Not run: 
pPerf <- xPredictROCR(prediction, GSP, GSN)

## End(Not run)

Pi documentation built on Nov. 26, 2020, 2:01 a.m.