DESP_SimpleOut: Detection of simple outliers
In DESP: Estimation of Diagonal Elements of Sparse Precision-Matrices

Description Usage Arguments Details Value Author(s)

This function aims to detect outliers characterized by a very large Euclidean norm.

1	DESP_SimpleOut(ve, method='MAD', iqr.mult=1.5, mad.mult=2.5, mad.constant=1.4826)

`ve`	Vector of Euclidean norms.
`method`	The method to be used : either based on interquantile range (IQR), or median absolute deviation (MAD, chosen by default).
`iqr.mult`	The multiplicative factor associated to IQR.
`mad.mult`	The multiplicative factor associated to MAD.
`mad.constant`	The scale constant associated to MAD.

Denoting the first quartile by Q1, the third one by Q3 and the median by Q2, if the method 'IQR' is chosen, any entry of ve above Q3 + iqr.mult * (Q3-Q1) is considered as an outlier. If the method 'MAD' is chosen, any entry of ve above Q2 + mad.mult * MAD is considered as an outlier.

This function returns the positions of the observations detected as outliers in ve.

Arnak Dalalyan and Samuel Balmand.

DESP documentation built on May 29, 2017, 9:27 p.m.