Description Usage Arguments Details Value Author(s) Examples

Wrapper for the detection of sample outliers by computation of Mahalanobis distances using
`cov.rob`

from package MASS. The (robust) square root distance from
the center is displayed alongside a 2D mapping of the data and its confidence
ellipse.

1 2 |

`x` |
A data.frame or matrix. |

`method` |
The method to be used: -
`mve` . Minimum volume ellipsoid. -
`mcd` . Minimum covariance determinant. -
`classical` . Classical product-moment.
For details, see |

`conf.level` |
The confidence level for controlling the cutoff of Mahalanobis distances. |

`dimen` |
Dimensions used to plot tolerance ellipse and the data points alongside these two dimensions. |

`tol` |
The tolerance to be used for computing Mahalanobis distances
(see |

`plotting` |
A logical value. If |

If the number of samples is `n`

and number of variables in a sample is
`p`

, the data set must be `n > p + 1`

. In this case, PCA can be
used to produce fewer directions of uncorrelated dimensions that explain
different dimensions in the data. Due to the inherent difficulties in
defining outliers, inclusion of the first few dimensions only is almost
always sufficient to compute Mahalanobis distances. However in more complex
designs implicating various factors and/or multiple levels, different
contributions to the overall variation modelled by PCA may be confounded in
such a reduced space. In such situation, the initial dataset must be
decomposed into smaller problems to relate potential outlying
behaviour.

A list with components:

`outlier` |
List of outliers detected. |

`conf.level` |
Confidence level used. |

`mah.dist` |
Mahalanobis distances of each data sample. |

`cutoff` |
Cutoff of Mahalanobis distances for outliers detection. |

Wanchang Lin [email protected] and David Enot [email protected]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | ```
## load abr1
data(abr1)
y <- factor(abr1$fact$class)
x <- preproc(abr1$pos , y=y, method=c("log10","TICnorm"),add=1)[,110:1000]
## Select classes 1 and 2
tmp <- dat.sel(x, y, choices=c("1","2"))
dat <- tmp$dat[[1]]
ind <- tmp$cl[[1]]
## dimension reduction by PCA
x <- prcomp(dat,scale=FALSE)$x
## perform and plot outlier detection using classical Mahalanobis distance
## on the first 2 PCA dimensions
res <- outl.det(x[,c(1,2)], method="classical",dimen=c(1,2),
conf.level = 0.975)
``` |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.