Table 4 Main characteristics of employed methods to assess an ISN’s significance.

LOO-ISN

TRUE

TRUE

Rep: number of repetition

Multivariate edge normality

TRUE

A null distribution is computed based on bootstrap

resampling assuming Normality—LOO procedure—

Aggregation method: sum of the absolute difference

MultiLOO-ISN

TRUE

TRUE

Rep: number of repetition

Multivariate edge normality

TRUE

As above—only difference the aggregation is

non-linear: maximum deviation from the null in

all edges

kNN

TRUE

TRUE

\(k_{min}\), \(k_{max}\) neighbour

/

FALSE

The distances found in kNN with k = \(k_{min}\) to

k = \(k_{max}\) are averaged to create outlier scores

Optics

TRUE

TRUE

MinPts: number of neighbour

/

FALSE

Outlier score is computed via a radius distance

of core and board points

Spoutlier

TRUE

TRUE

s: number of references

/

FALSE

The outlier score is calculated as the minimum

within a small set of references observations

kNN based

SSN-m

TRUE

FALSE

/

Edge normality

TRUE

The p-value is calculated as a transformation of

the difference between \(w^\alpha\) and \(w^{\alpha -q}\)

Cook distance

FALSE

TRUE

/

LM assumptions

FALSE

The outlier value is calculated aggregating cook’s

distance in every individual trying to predict one

of the edge weights in the modulus.