getCoords

Overview

This vignette describes the usage of the getCoords parameter in the pandemonium app.

This is an input for a named list of functions that can be used to calculate coordinates. The names of the list will be used as options for the coordinates selector in the GUI.

coordinates are a way to put each variable on equal footing for distance calculations. One simple coordinate function is normalising each variable by subtracting the mean and dividing by the standard deviation. Other forms are also possible and so we allow for the user to select what functions to pass to the app and which to use in the app.

Provided coordinate functions

normcoords

This uses scale to center and scale the data.

pullCoords

Chi-Squared Loss Function Coordinates

pullCoordsNoCov

Generic Loss Function Coordinates

rawCoords

Returns the data frame passed to it. This should only be used where data is already passed as coordinates or coordinate calculations are otherwise impossible.

userCoords

This is a closure of a coordinate function that returns the user defined coordinate matrix passed to the closure when called in the pandemonium call with pandemonium(df, getCoords=list(user=userCoords(coordMatrix))). This cannot be used with variables removed from a space.

Writing getCoords functions

Inputs

Input	Description
df	Data frame of raw values
cov	Covariance matrix of data frame
covinv	Inverse covariance matrix of data frame
exp	reference point

Ouput

The function should return a single array the same size of the input df and with the same column names as df.

Example

pullCoordsNoCov <- function(df, cov, exp, ...){

  n <- nrow(df)
  df <- as.matrix(df)
  nc <- ncol(df)
  coord_mat <- matrix(nrow = n, ncol = nc)

  for (i in 1:n){
    for (j in 1:nc){
      coord_mat[i, j] <- as.numeric((df[i, j] - exp$value[j]) / sqrt(cov[j, j]))
    }
  }
  colnames(coord_mat) <- colnames(df)
  return(coord_mat)
}

pandemonium(df,getCoords = list(pull = pullCoordsNoCov, normal = normCoords))