geomorph (version 4.1.1)

plotOutliers: Find potential outliers

Description

Function plots all specimens ordered by distance from the mean.

Usage

plotOutliers(A, groups = NULL, PC = NULL, inspect.outliers = FALSE)

Value

Function returns the landmark addresses of all specimens ordered as in the plot. If groups are used, function returns a list structure and a plot for each level in groups.

Arguments

A

A 3D array (p x k x n) containing Procrustes shape variables for a set of specimens

groups

An optional factor defining groups

PC

A single number or range of principal components that can be used to check for outliers. If NULL, all shape dimensions are used. This argument might be useful for investigating subtle but important aspects of shape that would not elucidate outliers in all dimensions.

inspect.outliers

A logical value indicating whether to plot outlier shape configurations as compared to the consensus

Author

Emma Sherratt, Antigoni Kaliontzopoulou, & Michael Collyer

Details

The function creates a plot of all specimens ordered by their distance from the mean shape. There is an option for using specified principal components (PC) of shape. If all shape dimensions are used then distances are equal to Procrustes distances or distances in the tangent space of shape space that resemble Procrustes distances, depending on whether the projection was performed with generalized Procrustes analysis (GPA). Once distances are calculated, a power-transformation is perfomed to normalize the distances. From these data, an upper limit is estimated following Tukey's box-plot rule, as \(Q_3 + 1.5 \times (Q_3 - Q_1)\), where \(Q\) refers to quartile. This upper limit is back-transformed to distance, and any distances greater than this limit are colored red. (Note: These shapes could be considered outliers but their red color does not mean they are necessarily outliers. )

The user may optionally also inspect the shapes of identified configurations as compared to the consensus, in order to identify potential digitization errors or other data issues. The addresses of all specimens are returned in the order displayed in the plot for further inspection by plotRefToTarget.

If the data have strong group structure and there is reasonable belief that the whole sample mean should not be used, then a factor defining the groups can be used.

See Also

gpagen

plotAllSpecimens

Examples

if (FALSE) {
data(plethodon)
# let's make some outliers
newland <- plethodon$land
newland[c(1,8),,2] <- newland[c(8,1),,2]
newland[c(3,11),,26] <- newland[c(11,3),,2]
Y <- gpagen(newland) # GPA
out <- plotOutliers(Y$coords) # function returns dimnames and address 
out
# of all specimens ordered
plotOutliers(Y$coords, inspect.outliers = TRUE) # function also produces 
# plots of identified outlier specimens compared to the mean shape

# example with groups
plotOutliers(Y$coords, groups = plethodon$species, 
inspect.outliers = TRUE)

# previous example using first three PCs of shape
plotOutliers(Y$coords, groups = plethodon$species, 
PC = 1:3, inspect.outliers = TRUE)

# previous example using just the first PC of shape
plotOutliers(Y$coords, groups = plethodon$species, 
PC = 1, inspect.outliers = TRUE)
 }