meandist.from.array: Tools for Working With Pairwise Distance Arrays

Description

meandist.from.array produces a mean distance matrix from an array of pairwise distances by locus, such as that produced by meandistance.matrix when all.distances=TRUE. find.na.dist finds missing distances in such an array, and find.na.dist.not.missing finds missing distances that aren't the result of missing genotypes.

Usage

meandist.from.array(distarray, samples = dimnames(distarray)[[2]],
loci = dimnames(distarray)[[1]])

find.na.dist(distarray, samples = dimnames(distarray)[[2]],
loci = dimnames(distarray)[[1]])

find.na.dist.not.missing(gendata, distarray, missing = -9,
samples = dimnames(distarray)[[2]], loci = dimnames(distarray)[[1]])

Arguments

distarray

A three-dimensional array of pairwise distances between samples, by locus. Loci are represented in the first dimension, and samples are represented in the second and third dimensions. Dimensions are named accordingly. Such an array is the first elem

samples

Character vector. Samples to analyze.

loci

Character vector. Loci to analyze.

gendata

A genotype object in the standard polysat format. Typically the genotype object that was used to produce distarray.

missing

The symbol used to represent missing data in gendata.

Value

meandist.from.array returns a matrix, with both rows and columns named by samples, of distances averaged across loci. find.na.dist and find.na.dist.not.missing both return data frames with three columns: Locus, Sample1, and Sample2. Each row represents the index in the array of an element containing NA.

Details

find.na.dist.not.missing is primarily intended to locate distances that were not calculated by Bruvo.distance because both genotypes had too many alleles (more than maxl). The user may wish to estimate these distances manually and fill them into the array, then recalculate the mean matrix using meandist.from.array.

Examples

Run this code

# set up the genotype data
samples <- paste("ind", 1:4, sep="")
samples
loci <- paste("loc", 1:3, sep="")
loci
testgen <- array(list(-9), dim = c(4,3), dimnames = list(samples,loci))
testgen[,"loc1"] <- list(c(-9), c(102,104), c(100,106,108,110,114),
                         c(102,104,106,110,112))
testgen[,"loc2"] <- list(c(77,79,83), c(79,85), c(-9), c(83,85,87,91))
testgen[,"loc3"] <- list(c(122,128), c(124,126,128,132), c(120,126),
                         c(124,128,130))

# look up which samples*loci have missing genotypes
find.missing.gen(testgen)

# get the three-dimensional distance array and the mean of the array
gendist <- meandistance.matrix(testgen, distmetric=Bruvo.distance,
                                maxl=4, all.distances=TRUE)
# look at the distances for loc1, where there is missing data and long genotypes
gendist[[1]]["loc1",,]

# look up all missing distances in the array
find.na.dist(gendist[[1]])

# look up just the missing distances that don't result from missing genotypes
find.na.dist.not.missing(testgen, gendist[[1]])

# Copy the array to edit the new copy
newDistArray <- gendist[[1]]
# calculate the distances that were NA from genotype lengths exceeding maxl
# (in reality, if this were too computationally intensive you might estimate
# it manually instead)
subDist <- Bruvo.distance(c(100,106,108,110,114), c(102,104,106,110,112))
subDist
# insert this distance into the correct positions
newDistArray["loc1","ind3","ind4"] <- subDist
newDistArray["loc1","ind4","ind3"] <- subDist
# calculate the new mean distance matrix
newMeanMatrix <- meandist.from.array(newDistArray)
# look at the difference between this matrix and the original.
newMeanMatrix
gendist[[2]]