Qscore: Data quality control
In monogeneaGM: Geometric Morphometric Analysis of Monogenean Anchors

Description Usage Arguments Details Value Author(s) References See Also Examples

This function checks data quality of anchor images by comparing the pairwise Euclidean distances of landmarks on the left and right anchors for both dorsal and ventral anchors.

1	Qscore(x, id.out = FALSE, tit = "", bound = 50)

`x`	a numeric vector generated from `pwdist`
`id.out`	if TRUE, allows the user to query data indices in the TMD plot
`tit`	title for the plot
`bound`	magnitude of upper and lower bounds on the y-axis

The quality control method implemented here is based on consideration of the difference (M) and average (A) of pairwise Euclidean distances between the left and right anchors. Let the residual be defined as the deviation of M from 0. Assuming that left and right anchors are more or less symmetric, good quality data should have squared residuals that are small on average and show more or less random deviation from 0 along A. The latter corresponds to slope that is near 0 in a linear regression of M against A. The result is visualized using the Tukey Mean-Difference (TMD) plot (also known as the Bland-Altman plot; see Bland & Altman (1986)), in which mean M and the 95 % limits of agreement (within 2 standard deviations from mean M) are indicated as dashed horizontal lines. Good quality data have small values of sqasr and slope, resulting in high Q scores, and vice versa.

A matrix with three columns:

`sqasr`	the square root of average squared residuals
`slope`	slope of regressing M against A, multiplied by 100 to be on the same scale as average squared residuals
`Q`	quality score, which is given by Q=100[10^(-sqrt(sqasr^2+slope^2)/10)]

Tsung Fei Khang tfkhang@um.edu.my

Bland JM, Altman DG. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 327:307-310.

Khang TF, Soo OYM, Tan WB, Lim LHS. (2016). Monogenean anchor morphometry: systematic value, phylogenetic signal, and evolution. PeerJ 4:e1668.

pwdist, boxplotSort

library(cluster)

data(ligophorus_tpsdata)
data(spcolmap)

#A low quality specimen
Qscore(pwdist(ligophorus_tpsdata$bantingensis[[5]],average=FALSE))
#A high quality specimen
Qscore(pwdist(ligophorus_tpsdata$bantingensis[[11]],average=FALSE))

#Useful diagnostic plots
Qmat <- vector("list",length(ligophorus_tpsdata))
for(i in 1:13){
Qmat[[i]] <- do.call(rbind,lapply(ligophorus_tpsdata[[i]],
function(k) Qscore(pwdist(k, average=FALSE))))
rownames(Qmat[[i]]) <- rep(names(ligophorus_tpsdata)[i],nrow(Qmat[[i]]))
}
names(Qmat) <- names(ligophorus_tpsdata)

#Box plot for quality score by species, sorted using descending median quality score
Q <- lapply(Qmat, function(k) k[,3])
boxplotSort(Q, italic=TRUE, ylab="Quality score", df=1)