predict_performance: Predicting performances of assemblages by only knowing their...

In functClust: Functional Clustering of Redundant Components of a System

Description

Takes a vector fct of assembly performances over several experiments and returns a vector of performances predicted as the mean performances of assemblages that share the same assembly motif.

Assembly motifs are labelled in the vector assMotif. Experiments are labelled in the vector xpr. Modelling options are indicated in opt.mean and opt.model. Occurrence matrix mOccur is used if opt.model = "byelt". Cross-validation is leave-one-out or jackknifesi

Usage

predict_performance(appFct, appMotifs, appOccur,
            supMotifs, supOccur,
            opt.mean = "amean",
            opt.model  = "bymot"  )

Arguments

appFct	a vector of numeric values (assembly properties).
appMotifs	a vector of labels of length(fct) (assembly motifs).
appOccur	a matrix of occurrence (occurrence of components). Its first dimension equals to length(fct). Its second dimension equals to the number of components.
supMotifs	a vector of labels of length(fct) (assembly motifs).
supOccur	a matrix of occurrence (occurrence of components). Its first dimension equals to length(fct). Its second dimension equals to the number of components.
opt.mean	equal to "amean" (by default) or "gmean".
opt.model	equal to "bymot" (by default) or "byelt".

Details

Prediction is computed using arithmetic mean amean by motif bymot in a whole (WITHOUT taking into account species contribution). The components belonging to a same motif are divided into jack[2] subsets of jack[1] components. Prediction is computed by excluding jack[1] components, of which the component to predict. If the total number of components belonging to the motif is lower than jack[1]*jack[2], prediction is computed by Leave-One-Out (LOO).

Value

Return the arithmetic mean of a vector, as standard mean function.

See Also

calibrate_byminrss
validate_using_cross_validation

functClust documentation built on Dec. 2, 2020, 5:06 p.m.