| validate {minet} | R Documentation |
validate compares the infered network to the true underlying network for several threshold values
and appends the resulting confusion matrices to the returned object.
validate(inet,tnet,steps=50,thresholds=pretty(inet,steps))
inet |
This is the infered network, a data.frame or matrix obtained by one of the functions minet, aracne, clr or mrnet . |
tnet |
The true underlying network. This network must have the same size and variable names as inet. |
steps |
The number of threshold values to be used in the validation process - see details. |
thresholds |
The vector of thresholds to be used. By default it is a sequence of about steps+1 equally spaced intervals. |
For each of the steps threshold values T, the edges whose weight are (strictly)
below T are eliminated. All the other edges will have a weight 1.
Thus for each threshold, we obtain a boolean network from the infered network. This
network is compared to the true underlying network, tnet, in order to compute a
confusion (adjacency) matrix.
All the confusion matrices, obtained with different threshold values, are appended to the
returned object.
In the end the validate function returns a data.frame containing steps+1 confusion matrices.
validate returns a data.frame whith four columns named thrsh, tp, fp, fn. These values are
computed for each of the steps thresholds. Thus each row of the returned object contains
the confusion matrix for a different threshold.
data(syn.data) data(syn.net) inf.net <- mrnet(build.mim(syn.data, estimator="spearman")) table <- validate( inf.net, syn.net, steps=100 ) table <- validate( inf.net, syn.net, steps=1, thresholds=0.5 )