ILDSR2: Compute R2 for Interlandmark Distances

In Morpho: Calculations and Visualisations Related to Geometric Morphometrics

Compute R2 for Interlandmark Distances

Description

Compute R2 for Interlandmark Distances explaining between group differences

Usage

ILDSR2(
  x,
  groups,
  R2tol = 0.9,
  autocluster = TRUE,
  gap = TRUE,
  bg.rounds = 999,
  wg.rounds = 999,
  which = 1:2,
  reference = NULL,
  target = NULL,
  mc.cores = 1,
  plot = FALSE,
  silent = FALSE,
  ...
)

Arguments

x	array containing landmarks
groups	vector containing group assignments or a numeric covariate. For groups with more than two levels, a pair of needs to be specified using which
R2tol	numeric: upper percentile for ILD R2 in relation to factor or in case autocluster=TRUE, the minimum quantile allowed.
autocluster	logical: if TRUE, the function is trying to find a cluster with the highest R2-values. In this case R2tol is used as the quantile the cluster is allowed to occupy.
gap	logical: if TRUE, the largest gap in the R2 distribution is sought. If FALSE, a clustering procedure is used.
bg.rounds	numeric: number of permutation rounds to assess between group differences
wg.rounds	numeric: number of rounds to assess noise within groups by bootstrapping.
which	integer (optional): in case the factor levels are > 2 this determins which factorlevels to use
reference	matrix containing start config landmarks. If NULL, it will be computed as mean for group 1.
target	matrix containing target config landmarks. If NULL, it will be computed as mean for group 2.
mc.cores	integer: number of cores to use for permutation tests.
plot	logical: if TRUE show graphical output of steps involved
silent	logical: suppress console output
...	additional parameters for internal use only.

Value

A list containing:

relevantILDs	containing landmark information with the highest R2-values
allR2	vector with ILD specific R2-values, sorted decreasingly
reftarILDS	matrix with columns containing ILDs for reference and target shapes
sampleILD	matrix containing ILDs of entire sample
R2tol	R2-threshold used
reference	reference used
target	target used
bg.test	result from between-group testing
confR2	confidence for relevant ILDs from bootstrapping
x	Procrustes superimposed raw data (without scaling)

Examples

require(Morpho)
data(boneData)
proc <- procSym(boneLM)
groups <- name2factor(boneLM,which=3)
ilds <- ILDSR2(proc$rotated,groups,plot=TRUE,wg.rounds=99,mc.cores=1)
if (interactive())
visualize(ilds)
## use covariate
## Not run: 
ildsLM <- ILDSR2(proc$rotated,groups=proc$size,plot=TRUE,wg.rounds=99,mc.cores=1)
if (interactive())
visualize(ildsLM)

## End(Not run)

## 2D Case with size as predictor
require(shapes)
require(Morpho)
gor.dat <- bindArr(gorf.dat,gorm.dat,along=3)
procg <- procSym(gor.dat)
ildsg <- ILDSR2(procg$rotated,procg$size,plot=FALSE,bg.rounds=999,wg.rounds=99,
                mc.cores=1,autocluster=TRUE,R2tol=.8)

Morpho documentation built on Sept. 9, 2025, 5:46 p.m.