phylter (version 0.9.12)

detect.outliers: Detection of outliers in 1D and 2D data

Description

Functions to detect outliers, in matrices or in arrays.

Usage

detect.outliers(X, k = 3, test.island = TRUE, normalizeby = "row")
detect.outliers.array(arr, nbsp, k = 3)

Value

detect.outliers: A matrix with outliers detected in the 2D matric. Each row x contains the gene (x[1]) where the species (x[2]) is an outlier.

detect.outliers.array: An array listing the outliers detected (if any)

Arguments

X

2D matrix (gene x species) obtained with the Dist2WR() function.

k

strength of outlier detection. High values (typically 3) correspond to less outliers detected than with lower ones (e.g. 1.5).

test.island

should islands of outliers be treated as such. If TRUE (the default), only the highest value in an island of outliers is removed. This prevents erroneous outliers due to hicthiking effect to be removed.

normalizeby

Should the 2D matrix be normalized prior to outlier detection, and how. Can be "row" (the default),"col" or "none". Normalization is done by dividing columns or rows by their median.

arr

Array of values, typically the weight of each gene matrix (alpha values).

nbsp

Number of species in the analysis

Functions

• detect.outliers(): detect outliers in 2D matrix

• detect.outliers.array(): detects outliers in 1D array

Details

These functions detect outliers either in matrices or in arrays. For the method to be adapted to skewed data, as is the case here, the outlier detection method used is the adjusted Tukey proposed by Hubert and Vandervieren (2008).

References

de Vienne D.M., Ollier S. et Aguileta G. (2012) Phylo-MCOA: A Fast and Efficient Method to Detect Outlier Genes and Species in Phylogenomics Using Multiple Co-inertia Analysis. Molecular Biology and Evolution 29 : 1587-1598.

Hubert, M. and Vandervieren, E. (2008). An adjusted boxplot for skewed distributions. Computational Statistics and Data Analysis, 52, 5186-5201.

Examples

# Get the initial gene x species matrix
# from the carnivora dataset
data(carnivora) 
mat <- phylter(carnivora, InitialOnly = TRUE, parallel = FALSE)$WR

# detect outliers in this matrix
outliers<-detect.outliers(mat)
outliers$cells # matrix where each row represents one cell in the input matrix