README.md

In matthew-seth-smith/replicateOutliers: Outlier Detection Methods for Replicated Data

replicateOutliers

This package provides functions for outlier detection in the setting of replicated data. Assuming we have multiple independent measurements (called "replicates") for each data point, we can find which sets of measurements are so different from each other that we can call them outliers. We use the absolute difference and coefficient of variation (Zeta, a kind of relative difference) for each pair of replicates to make our determinations.

The function outlier_DZ returns a numeric identifier for the outlier status ( for non-outlier, for large but not , and for outlier). The other main functions are q_exp_joint_DZ, q_exp_marg_DZ, q_gg_joint_DZ, and q_gg_marg_DZ. Their outputs are probabilities called -values that we determine using joint or marginal probability distributions.

Please see {CITE PAPER} for further details and the derivations of these methods. Also please check the vignette HTML file in the inst/doc directory.

Currently, outlier_DZ does not work, because it depends on the extremevalues package for the getOutliersI function. This package, in turn, depends on the tcltk package, which is now part of base R. The package extremevalues should then be able to run without importing tcltk, but it does not work. If you are able to get extremevalues to work locally, the code for outlier_DZ is in the vignette explaining this method.

To install this package, please install and load the devtools package and use the command

install_github("matthew-seth-smith/replicateOutliers")

Travis-CI is throwing errors for this build (as does loading this function on another computer) because of conflicts between the gompertz functions used by the packages VGAM and flexsurv. I do not believe this package uses the Gompertz Distribution, so this should not be a problem.

This project was supported by the Undergraduate Summer Research Fellowship at Fox Chase Cancer Center during the summer of 2018.

matthew-seth-smith/replicateOutliers documentation built on Jan. 24, 2020, 9:34 p.m.