paulbarber/msawrapper

This package unifies the exploration and use of multivariate survival analysis methods for time to event and ordinal class data.

msawrapper's Issues

SPS processing confusion with previous runs

SPS creates new folders for each run. We currently perform prediction from the first run, not the last.
Maybe check for previous runs and adjust the ini file appropriately.
Some SPS plot functions depend on getting data from the logfile. This should happen when the model is built, not when result is plotted, and should extract the last entry (seems to extract all entries for the optimal number of covariates?). Try to remove all dependence on the log file post model building.

Use signature on test data

Is there a nice way to test a signature on some new/test data, and measure performance?
Rank corr, and AUC for binary outcomes.

Need for Strong Outcome correlations function

Similar to strongDataCorrelations

rowLabels lost by reduceByCorrelationWithinGroups

The new msaWrapper produced by reduceByCorrelationWithinGroups() does not have the low labels of the old msaWrapper.

Need a function to print what the signature is.

This should report the covars and weights.
Also, plot the whisker plot of HR?