This version of chemmodlab
has all of the
functionality mentioned in the following paper:
https://arxiv.org/abs/1912.09526
Added function PerfCurveBands
which constructs a
confidence band for a recall curve.
Added function PerfCurveTest
which performs a
hypothesis test for the difference between two performance
curves.
Added function HitEnrich
which plots hit enrichment
curves and confidence bands.
Added function HitEnrichDiff
which plots the
difference between hit enrichment curves and confidence bands.
This version of chemmodlab
has all of the
functionality mentioned in the following paper:
https://jcheminf.biomedcentral.com/articles/10.1186/s13321-018-0309-4
Added function ApplicabilityDomain
which uses a
Hotelling T2 control chart to identify outliers in an external data set
for which predictions are desired, hence identifying observations whose
model predictions may be considered extrapolations.
ModelTrain
now has a S3 method which accepts
molecule objects that are created by the package rcdk. rcdk supports
most of the widely used chemical file formats (SMILES, SDF, InChI, Mol2,
CML, etc). When molecules are provided to ModelTrain
, the
names of predefined descriptor sets and/or fingerprints must also be
provided.y
CombineSplits()
when ppv was used as the model
performance measure, it at times evaluated to NA because there were no
predicted positives for a model. Performance()
may be used
to find these cases. The ppv for these splits is now imputed using the
mean of the other splits that are not NA. If there are no splits that
are not NA, CombineSplits()
returns an errorFirst feature complete release of chemmodlab on CRAN