Pairwise distance CPG, polarization, or segregation measure

A Ratio-Polarization Measure.R

require(cluster)

Polarizer <- function(group, data){

  if((mean(is.na(group)) > 0.99) | length(group) <= 1){
    Output <- c(NA, NA)
    } else {

  SameGroup <- try(as.matrix(daisy(data.frame(group))) == 0)
  diag(SameGroup) <- NA
  SameGroup[lower.tri(SameGroup)] <- NA

  PairwiseDistance <- as.matrix(dist(data))
  PairwiseDistance[lower.tri(PairwiseDistance)] <- NA

  WithinDistances <- PairwiseDistance[SameGroup == 1]
  BetweenDistances <- PairwiseDistance[SameGroup != 1]

  WithinMean <- mean(WithinDistances, na.rm = T)
  BetweenMean <- mean(BetweenDistances, na.rm = T)
  WithinSE <- sd(WithinDistances, na.rm = T) / sqrt(sum(!is.na(WithinDistances)))
  BetweenSE <- sd(BetweenDistances, na.rm = T) / sqrt(sum(!is.na(BetweenDistances)))
  MeanRatio <- BetweenMean / WithinMean
  LogMeanRatio <- log(MeanRatio, base = 10)

  RatioSE <- MeanRatio * sqrt((WithinSE / WithinMean) ^ 2 + (BetweenSE / BetweenMean) ^ 2)

  # This returns a logged Estimate, but an unlogged Std. Err:
  Output <- c(LogMeanRatio, RatioSE)
    }
  names(Output) <- c("Coefficient", "SE")
  return(Output)
  }

Pairwise Distance CPG.R

Polarizer <- function(group, data){
  SameGroup <- as.matrix(dist(as.numeric(as.factor(group)))) == 0
  diag(SameGroup) <- NA
  SameGroup[lower.tri(SameGroup)] <- NA

  PairwiseDistance <- as.matrix(dist(data))
  PairwiseDistance[lower.tri(PairwiseDistance)] <- NA
  PairwiseDistance <- PairwiseDistance / max(PairwiseDistance, na.rm = T)

  WithinDistances <- PairwiseDistance[SameGroup == 1]
  BetweenDistances <- PairwiseDistance[SameGroup != 1]

  WithinMean <- mean(WithinDistances, na.rm = T)
  BetweenMean <- mean(BetweenDistances, na.rm = T)
  WithinSE <- sd(WithinDistances, na.rm = T) / sqrt(sum(!is.na(WithinDistances)))
  BetweenSE <- sd(BetweenDistances, na.rm = T) / sqrt(sum(!is.na(BetweenDistances)))
  MeanDifference <- BetweenMean - WithinMean
  MeanDifferenceSE <- sqrt(WithinSE ^ 2 + BetweenSE ^ 2)
  Output <- c(MeanDifference, MeanDifferenceSE)
  names(Output) <- c("Coefficient", "SE")
  return(Output)
  }

Spatial Segregation.R

# Normalized and logistic'd:

Polarizer <- function(group, data){
  Normalize <- function(x){(x - mean(x, na.rm = T)) / sd(x, na.rm = T)}
  Logistic <- function(x){exp(x) / (1 + exp(x))}

  SameGroup <- as.matrix(dist(as.numeric(as.factor(group)))) == 0
  diag(SameGroup) <- NA
  SameGroup[lower.tri(SameGroup)] <- NA

  PairwiseDistance <- as.matrix(dist(data))
  PairwiseDistance[lower.tri(PairwiseDistance)] <- NA
  PairwiseDistance <- Normalize(PairwiseDistance)
  PairwiseDistance <- Logistic(PairwiseDistance)

  WithinDistances <- PairwiseDistance[SameGroup == 1]
  BetweenDistances <- PairwiseDistance[SameGroup != 1]

  WithinMean <- mean(WithinDistances, na.rm = T)
  BetweenMean <- mean(BetweenDistances, na.rm = T)
  WithinSE <- sd(WithinDistances, na.rm = T) / sqrt(sum(!is.na(WithinDistances)))
  BetweenSE <- sd(BetweenDistances, na.rm = T) / sqrt(sum(!is.na(BetweenDistances)))
  MeanDifference <- BetweenMean - WithinMean
  MeanDifferenceSE <- sqrt(WithinSE ^ 2 + BetweenSE ^ 2)
  Output <- c(MeanDifference, MeanDifferenceSE)
  names(Output) <- c("Coefficient", "SE")
  return(Output)
  }

Use.R

# Some sample data:

N <- 100
Groups <- sort(sample(LETTERS[1:3], N, replace = T))
Locations <- matrix(rnorm(3*100), ncol = 3)
Locations <- Locations[order(Locations[, 1]), ]
plot(Locations, type = "n")
text(Locations[, 1:2], Groups, cex = exp(Locations[, 3]) / (1+exp(Locations[, 3])))

# Test:

Polarizer(Groups, Locations)