R Graphical Manual

Last data update: 2014.03.03

R: Feature selection method

HD2013SGIselectByStability

R Documentation

Feature selection method

A function to select features that are most stable across replicated experiments

HD2013SGIselectByStability(subsample, preselect = "count", Rdim = 40, verbose = TRUE)

`subsample`	This is the input data. Usually a subsample of the complete screen is enough to select the non-redundant features. `subsample` is a list with three elements: `D` is a three-dimensional array with dimensions samples x features x replicates. As samples usually 1000 to 5000 experiments are randomly selected. The function needs two replicates.
`preselect`	Names of the features that should be preselected, e.g. count is preselected in this screen, because of its biological interpretability and comparability to other viability-based genetic interaction screens.
`Rdim`	The maximum number of features to be selected.
`verbose`	If TRUE information about the progress and the quality of the selected features is printed.

(selected = selected, correlation = correlation, ratioPositive = ratioPositive, correlationAll = correlationAll)

`selected`	The names of the selected features in the order as selected.
`correlation`	The correlation of the residual features after fitting a linear function on the previously selected features. Correlations are in same order as `selected`.
`ratioPositive`	The fraction of positively correlated features among all candidate features in each step. In same order as `selected`.
`correlationAll`	The correlation of the residual features of all candidate features in each step of the selection process.