createTruthPlot: Create the "truth plot"
In PriceLab/FPML: Footprint Machine Learning

Description Usage Arguments Value

Create a "truth" plot, which plots HINT v. Wellington scores given a model and a threshold to use for determining a "positive"

1	createTruthPlot(model, modelType, seed, threshold, X_test, y_test, plotFile)

`model`	A model trained on the annotated dataset, either a linear or gradient-boosted model
`modelType`	Either "boosted" or "linear" to indicate the type of model provided
`seed`	One of "16", "20", or "both", denoting which seed is being used
`threshold`	A numeric from 0-1 indicating the cutoff for what constitutes a "positive" probability. This is an INCLUSIVE value; positives are greater than or equal to the threshold
`X_test`	The matrix of test data for predictors, generally created using "prepModelData"
`y_test`	The matrix of test data for response, generally created using "prepModelData"
`plotFile`	A location to use for creating the png plot file

The "truth data frame", containing the Wellington/HINT scores for each point in the test set, the predicted probability of a ChIPseq hit, the binary prediction based on that probability and the supplied threshold, the true ChIPseq hit value, and the corresponding label of TP/TN/FP/FN.

It also constructs and saves a plot of HINT vs Wellington scores, where both have been transformed via asinh() and the Wellington scores have been converted via absolute value. Points are coded by confusion matrix outcome.

PriceLab/FPML documentation built on May 28, 2019, 2:25 p.m.