Package 'GPAbin'

Biplot function

Description

Creates a multiple correspondence analysis (MCA) biplot

Usage

biplFig(
  missbp,
  Z.col = "#61223b",
  CLP.col = "#b79962",
  Z.pch = 19,
  CLP.pch = 15,
  Z.cex = 1.5,
  CLP.cex = 1.7,
  title = ""
)

Arguments

missbp	An object of class missbp obtained from preceding function missmi()
Z.col	Colour of sample coordinates
CLP.col	Colour of category level point coordinates
Z.pch	Plotting character of sample coordinates
CLP.pch	Plotting character of category level point coordinates
Z.cex	Size of plotting character for sample points
CLP.cex	Size of plotting character for category level point points
title	Title of the plot

Value

• If compdat = NULL in evalMeas, only a GPAbin biplot will be constructed.

• If a complete data set (compdat) was specified in evalMeas, two biplots will be constructed: (1) Complete MCA biplot and (2) GPAbin biplot.

Examples

data(implist)
missbp <- missmi(implist)|> DRT() |> GPAbin() |> biplFig()

---

Category level prediction

Description

Predicts category levels from an MCA based biplot using the distances between coordinates

Usage

CLPpred(CLPs = CLPs, Zs = Zs, p = p, n = n, lvls = lvls, datIN = datIN)

Arguments

CLPs	Category level point coordinates
Zs	Sample coordinates
p	Number of variables
n	Number of samples
lvls	Names of category levels
datIN	Input data from which CLPs and Zs are obtained

Value

predCL

Final predicted categorical data set

---

Complete data example

Description

Simulated data example

Format

A data frame with 1000 rows and 5 columns.

Details

V1

Variable 1

V1

Variable 2

V1

Variable 3

V1

Variable 4

V1

Variable 5

Source

Simulated data from a uniform distribution that is categorised into levels.

---

Dimension reduction

Description

Multiple correspondence analysis is performed on the multiple imputed datasets

Usage

DRT(missbp, method = c("MCA"))

Arguments

missbp	An object of class missbp obtained from preceding function missmi()
method	Select a dimension reduction technique. In the current version MCA is available.

Value

The missbp object is appended with the following objects:

Z	List of sample coordinates
CLP	List of category level point coordinates
lvls	List of category level names
m	Number of multiple imputations

See also missmi and impute.

Examples

data(implist)
missbp <- missmi(implist) |> DRT()

---

Evaluation measures when complete data is available

Description

Calculates measures of comparison based on distances between two configurations in two dimensions.

Usage

evalMeas(missbp, compdat = NULL)

Arguments

missbp	An object of class missbp obtained from preceding function missmi().
compdat	Complete data matrix representing the input data of missmi()

Value

The missbp object is appended with the following objects:

eval	Returns a data table with five evaluation measures: Procrustes Statistic (PS), Similarity Proportion (SP), Response Profile Recovery (RPR), Absolute Mean Bias (AMB), Root Mean Squared Bias (RMSB)
GPApred	A dataframe representing predicted categorical responses from the GPAbin biplot.
compPred	A dataframe representing predicted categorical responses from the complete MCA biplot.
compZs	Sample coordinates for the MCA biplot of the complete data.
compCLPs	CLPs for the MCA biplot of the complete data.
complvls	Names of the CLPs for the MCA biplot of the complete data.

See also missmi, impute, DRT and GPAbin.

For more detail, refer to Nienkemper-Swanepoel, J., le Roux, N. J., & Gardner-Lubbe, S. (2021). GPAbin: unifying visualizations of multiple imputations for missing values. Communications in Statistics - Simulation and Computation, 52(6), 2666–2685. https://doi.org/10.1080/03610918.2021.1914089.

Examples

data(compdat)
data(implist)
missbp <- missmi(implist) |> DRT() |> GPAbin() |> evalMeas(compdat=compdat)

---

ggBiplot function

Description

Creates a multiple correspondence analysis (MCA) biplot in ggplot.

Usage

ggbiplFig(
  missbp,
  Z.col = "#61223b",
  CLP.col = "#b79962",
  Z.pch = 19,
  CLP.pch = 15
)

Arguments

missbp	An object of class missbp obtained from preceding function missmi()
Z.col	Colour of sample coordinates
CLP.col	Colour of category level point coordinates
Z.pch	Plotting character of sample coordinates
CLP.pch	Plotting character of category level point coordinates

Value

• plotReturns a GPAbin ggplot biplot.

• plotCReturns an MCA ggplot biplot of the complete data set, if provided.

Examples

data(implist)
data(compdat)
missbp <- missmi(implist)|> DRT() |> GPAbin() |> 
evalMeas(compdat = compdat) |> ggbiplFig()

### GPAbin biplot

missbp$plot

### MCA biplot

missbp$plotC

---

Generalised Orthogonal Procrustes Analysis

Description

This function contains the OPA function to compare two configurations and the GPA function for multiple configuration comparisons

Usage

GPA(Xk, G.target = NULL, iter = 500, eps = 0.001)

Arguments

Xk	list containing the testee configurations which is updated on #each iteration
G.target	Target configuration. If not specified the centroid configuration will be used as the target
iter	Number of iterations allowed before convergence
eps	Threshold value for convergence of the alogrithm

Value

Xk.F	List containing the updated testee configurations
sk.F	Vector containing the final scaling factors
Qk.F	List containing the final rotation matrices
Gmat	Final target configuration
sum.sq	Final minimised sum of squared distance

---

Function to unify coordinates of multiple configurations

Description

Combines multiple configurations from dimension reduction solutions applied to multiple imputed data sets

Usage

GPAbin(missbp, G.target = NULL)

Arguments

missbp	An object of class missmi obtained from preceding function missmi()
G.target	Target configuration. If not specified the centroid configuration will be used as the target.

Value

The missbp object is appended with the following objects:

Z.GPA.list	List containing the sample coordinates for each MI after GPA
CLP.GPA.list	List containing the CLPs for each MI after GPA
G.target	Target configuration
Z.GPAbin	Sample coordinates for the GPAbin biplot
CLP.GPAbin	CLPs for the GPAbin biplot

See also missmi, impute and DRT.

For more detail, refer to Nienkemper-Swanepoel, J., le Roux, N. J., & Gardner-Lubbe, S. (2021). GPAbin: unifying visualizations of multiple imputations for missing values. Communications in Statistics - Simulation and Computation, 52(6), 2666–2685. https://doi.org/10.1080/03610918.2021.1914089.

Examples

data(implist)
missbp <- missmi(implist) |> DRT() |> GPAbin()

---

List of multiple imputed data sets

Description

Five multiple imputations of missdat

Format

List containing five multiple imputations of missdat. Each list item a data frame with 1000 rows and 5 columns.

Details

V1

Variable 1

V1

Variable 2

V1

Variable 3

V1

Variable 4

V1

Variable 5

Source

simulated example data imputed with mice::mice(missdat, m=5, method="polyreg", maxit=10, remove.collinear=FALSE, printFlag = FALSE)

---

Multiple imputation

Description

Choose between four available multiple imputation strategies in R.

Usage

impute(missbp, imp.method = c("MIMCA", "jomo", "DPMPM", "mice"), m = 5)

Arguments

missbp	An object of class missmi obtained from preceding function missmi().
imp.method	Select one of four imputation methods: MIMCA, jomo, DPMPM, mice
m	Number of multiple imputations

Value

The missbp object is appended with the following object:

dataimp

List of imputed data

See also MIMCA, jomo1cat, mi and mice.

Examples

data(missdat)
missbp <- missmi(missdat) |> impute(imp.method="DPMPM", m=5)

---

Missing data example

Description

compdat containing approximately 35% simulated missing values according to a missing at random (MAR) missing data mechanism

Format

A data frame with 1000 rows and 5 columns.

Details

V1

Variable 1

V1

Variable 2

V1

Variable 3

V1

Variable 4

V1

Variable 5

Source

Simulated data from a uniform distribution that is categorised into levels.

---

First step before constructing unified biplots

Description

This function produces a list of elements to be used when producing a GPAbin biplot.

Usage

missmi(data)

Arguments

data	input data frame or list

Value

X	The processed data
m	Number of multiple imputations applied
n	The number of samples
p	The number of variables
miss_pct	Percentage of missing values

Examples

data(missdat)
missbp <- missmi(missdat)
data(implist)
missbp <- missmi(implist)

---

Orthogonal Procrustes Analysis

Description

This function performs Orthogonal Procrustes Analysis on centred data

Usage

OPA(missbp, compdat, centring = TRUE, dim = "2D")

Arguments

missbp	An object of class missmi obtained from preceding function missmi()
compdat	Complete data set, only available for simulated data examples.
centring	Logical argument to apply centering, default is TRUE.
dim	Number of dimensions to use in final solutions (⁠2D⁠ or All available dimensions.)

Value

ProcStat	Procrustes Statistic
compZ	Sample coordinates representing the complete data set
compCLP	Category level point coordinates representing the complete data set
complvls	Category levels
compdat	Complete data set, only available for simulated data examples

---

Generic print function for objects of class missmi

Description

This function is used to print output when the missmi biplot object is created.

Usage

## S3 method for class 'missmi'
print(x, ...)

Arguments

x	an object of class missmi.
...	additional arguments.

Value

This function will not produce a return value, it is called for side effects.

Examples

data(missdat)
missbp <- missmi(missdat)
data(implist)
missbp <- missmi(implist)
print(missbp)

---