# Libraries to be used
# Libraries to be used
library(tidyverse)
library(mice) # Library to generate missing values and to use MICE technique
library(VIM) # Library to impute NAs with KNN
library(missForest) # Library to impute NAs with random forest
library(DT) # Library to have an interacting table
library(ggpubr) # Combine plots
library(Hmisc) # plot hist can be used to impute too
library(naniar) #visualization tool
library(reshape2) # to convert data from wide to long format
# Dataset Import and Initial Manipulation
# Load dataset
dataset <- read.csv("/Users/gabrielrivera/Documents/Maestria/Courses/5- Fall 2022/STA6257/Missing Data Practice/GlassdoorGenderPayGap.csv")
data <- dataset #%>% select(JobTitle, Gender, Age, Seniority)
data$Age <- as.numeric(data$Age)
data$PerfEval <- as.factor(data$PerfEval)
data$Seniority <- as.factor(data$Seniority)
data$BasePay <- as.numeric(data$BasePay)
data$Bonus <- as.numeric(data$Bonus)
# Generate Random Missing Values
# Generate MCAR Missing Values
dataMCAR <- ampute(data, prop=0.2, mech="MCAR")$amp # The ampute generates a bunch of unnecesary things and $amp has the data w NAs
# Get the location index of the NAs
index_MCAR <- as_tibble(which(is.na(dataMCAR), arr.ind=TRUE))
# Generate vector that contains the original data that got an NA
JobTitle_MCAR_NAs <- data$JobTitle [filter(index_MCAR, index_MCAR$col==1)$row]
Gender_MCAR_NAs <- data$Gender [filter(index_MCAR, index_MCAR$col==2)$row]
Age_MCAR_NAs <- data$Age [filter(index_MCAR, index_MCAR$col==3)$row]
PerfEval_MCAR_NAs <- data$PerfEval [filter(index_MCAR, index_MCAR$col==4)$row]
Education_MCAR_NAs <- data$Education[filter(index_MCAR, index_MCAR$col==5)$row]
Dept_MCAR_NAs <- data$Dept [filter(index_MCAR, index_MCAR$col==6)$row]
Seniority_MCAR_NAs <- data$Seniority[filter(index_MCAR, index_MCAR$col==7)$row]
BasePay_MCAR_NAs <- data$BasePay [filter(index_MCAR, index_MCAR$col==8)$row]
Bonus_MCAR_NAs <- data$Bonus [filter(index_MCAR, index_MCAR$col==9)$row]
# MAR and MNAR Pre-Processing
data$JobTitle <- as.factor(data$JobTitle)
data$Gender <- as.factor(data$Gender)
data$Education <- as.factor(data$Education)
data$Dept <- as.factor(data$Dept)
JobTitle_levels <- levels(data$JobTitle)
Gender_levels <- levels(data$Gender)
Education_levels <- levels(data$Education)
Dept_levels <- levels(data$Dept)
# Generate MAR Missing Values
dataMAR <- ampute(data, prop=0.2, mech="MAR")$amp
# Substitute MAR numeric values back to character
dataMAR$JobTitle <- as.character(JobTitle_levels[dataMAR$JobTitle])
dataMAR$Gender <- as.character(Gender_levels[dataMAR$Gender])
dataMAR$Education <- as.character(Education_levels[dataMAR$Education])
dataMAR$Dept <- as.character(Dept_levels[dataMAR$Dept])
dataMAR$PerfEval <- as.factor(dataMAR$PerfEval)
dataMAR$Seniority <- as.factor(dataMAR$Seniority)
# Get the location index of the NAs
index_MAR <- as_tibble(which(is.na(dataMAR), arr.ind=TRUE))
# Generate vector that contains the original data that got an NA
JobTitle_MAR_NAs <- data$JobTitle [filter(index_MAR, index_MAR$col==1)$row]
Gender_MAR_NAs <- data$Gender [filter(index_MAR, index_MAR$col==2)$row]
Age_MAR_NAs <- data$Age [filter(index_MAR, index_MAR$col==3)$row]
PerfEval_MAR_NAs <- data$PerfEval [filter(index_MAR, index_MAR$col==4)$row]
Education_MAR_NAs <- data$Education[filter(index_MAR, index_MAR$col==5)$row]
Dept_MAR_NAs <- data$Dept [filter(index_MAR, index_MAR$col==6)$row]
Seniority_MAR_NAs <- data$Seniority[filter(index_MAR, index_MAR$col==7)$row]
BasePay_MAR_NAs <- data$BasePay [filter(index_MAR, index_MAR$col==8)$row]
Bonus_MAR_NAs <- data$Bonus [filter(index_MAR, index_MAR$col==9)$row]
# Generate MNAR Missing Values
dataMNAR <- ampute(data, prop=0.2, mech="MNAR")$amp
# Substitute MNAR numeric values back to character
dataMNAR$JobTitle <- as.character(JobTitle_levels[dataMNAR$JobTitle])
dataMNAR$Gender <- as.character(Gender_levels[dataMNAR$Gender])
dataMNAR$Education <- as.character(Education_levels[dataMNAR$Education])
dataMNAR$Dept <- as.character(Dept_levels[dataMNAR$Dept])
dataMNAR$PerfEval <- as.factor(dataMNAR$PerfEval)
dataMNAR$Seniority <- as.factor(dataMNAR$Seniority)
# Get the location index of the NAs
index_MNAR <- as_tibble(which(is.na(dataMNAR), arr.ind=TRUE))
# Generate vector that contains the original data that got an NA
JobTitle_MNAR_NAs <- data$JobTitle [filter(index_MNAR, index_MNAR$col==1)$row]
Gender_MNAR_NAs <- data$Gender [filter(index_MNAR, index_MNAR$col==2)$row]
Age_MNAR_NAs <- data$Age [filter(index_MNAR, index_MNAR$col==3)$row]
PerfEval_MNAR_NAs <- data$PerfEval [filter(index_MNAR, index_MNAR$col==4)$row]
Education_MNAR_NAs <- data$Education[filter(index_MNAR, index_MNAR$col==5)$row]
Dept_MNAR_NAs <- data$Dept [filter(index_MNAR, index_MNAR$col==6)$row]
Seniority_MNAR_NAs <- data$Seniority[filter(index_MNAR, index_MNAR$col==7)$row]
BasePay_MNAR_NAs <- data$BasePay [filter(index_MNAR, index_MNAR$col==8)$row]
Bonus_MNAR_NAs <- data$Bonus [filter(index_MNAR, index_MNAR$col==9)$row]
# Single Imputation (SI) using Mean and Most Frequent (MF) value
# Impute MCAR NAs with mean for continuous variables and most frequent for categorical variables
MCAR_SI <- tibble( JobTitle = as.character(names(which.max(table(dataMCAR$JobTitle)))),
Gender = as.character(names(which.max(table(dataMCAR$Gender)))),
Age = mean(dataMCAR$Age,na.rm = TRUE),
PerfEval = as.character(names(which.max(table(dataMCAR$PerfEval)))),
Education = as.character(names(which.max(table(dataMCAR$Education)))),
Dept = as.character(names(which.max(table(dataMCAR$Dept)))),
Seniority = as.factor(names(which.max(table(dataMCAR$Seniority)))),
BasePay = mean(dataMCAR$BasePay,na.rm = TRUE),
Bonus = mean(dataMCAR$Bonus,na.rm = TRUE))
dataMCAR_SI <- dataMCAR
dataMCAR_SI$JobTitle[is.na(dataMCAR_SI$JobTitle)] <- MCAR_SI$JobTitle
dataMCAR_SI$Gender[is.na(dataMCAR_SI$Gender)] <- MCAR_SI$Gender
dataMCAR_SI$Age[is.na(dataMCAR_SI$Age)] <- MCAR_SI$Age
dataMCAR_SI$PerfEval[is.na(dataMCAR_SI$PerfEval)] <- MCAR_SI$PerfEval
dataMCAR_SI$Education[is.na(dataMCAR_SI$Education)] <- MCAR_SI$Education
dataMCAR_SI$Dept[is.na(dataMCAR_SI$Dept)] <- MCAR_SI$Dept
dataMCAR_SI$Seniority[is.na(dataMCAR_SI$Seniority)] <- MCAR_SI$Seniority
dataMCAR_SI$BasePay[is.na(dataMCAR_SI$BasePay)] <- MCAR_SI$BasePay
dataMCAR_SI$Bonus[is.na(dataMCAR_SI$Bonus)] <- MCAR_SI$Bonus
# Impute MAR NAs with mean for continuous variables and most frequent for categorical variables
MAR_SI <- tibble( JobTitle = as.character(names(which.max(table(dataMAR$JobTitle)))),
Gender = as.character(names(which.max(table(dataMAR$Gender)))),
Age = mean(dataMAR$Age,na.rm = TRUE),
PerfEval = as.character(names(which.max(table(dataMAR$PerfEval)))),
Education = as.character(names(which.max(table(dataMAR$Education)))),
Dept = as.character(names(which.max(table(dataMAR$Dept)))),
Seniority = as.factor(names(which.max(table(dataMAR$Seniority)))),
BasePay = mean(dataMAR$BasePay,na.rm = TRUE),
Bonus = mean(dataMAR$Bonus,na.rm = TRUE))
dataMAR_SI <- dataMAR
dataMAR_SI$JobTitle[is.na(dataMAR_SI$JobTitle)] <- MAR_SI$JobTitle
dataMAR_SI$Gender[is.na(dataMAR_SI$Gender)] <- MAR_SI$Gender
dataMAR_SI$Age[is.na(dataMAR_SI$Age)] <- MAR_SI$Age
dataMAR_SI$PerfEval[is.na(dataMAR_SI$PerfEval)] <- MAR_SI$PerfEval
dataMAR_SI$Education[is.na(dataMAR_SI$Education)] <- MAR_SI$Education
dataMAR_SI$Dept[is.na(dataMAR_SI$Dept)] <- MAR_SI$Dept
dataMAR_SI$Seniority[is.na(dataMAR_SI$Seniority)] <- MAR_SI$Seniority
dataMAR_SI$BasePay[is.na(dataMAR_SI$BasePay)] <- MAR_SI$BasePay
dataMAR_SI$Bonus[is.na(dataMAR_SI$Bonus)] <- MAR_SI$Bonus
# Impute MNAR NAs with mean for continuous variables and most frequent for categorical variables
MNAR_SI <- tibble( JobTitle = as.character(names(which.max(table(dataMNAR$JobTitle)))),
Gender = as.character(names(which.max(table(dataMNAR$Gender)))),
Age = mean(dataMNAR$Age,na.rm = TRUE),
PerfEval = as.character(names(which.max(table(dataMNAR$PerfEval)))),
Education = as.character(names(which.max(table(dataMNAR$Education)))),
Dept = as.character(names(which.max(table(dataMNAR$Dept)))),
Seniority = as.factor(names(which.max(table(dataMNAR$Seniority)))),
BasePay = mean(dataMNAR$BasePay,na.rm = TRUE),
Bonus = mean(dataMNAR$Bonus,na.rm = TRUE))
dataMNAR_SI <- dataMNAR
dataMNAR_SI$JobTitle[is.na(dataMNAR_SI$JobTitle)] <- MNAR_SI$JobTitle
dataMNAR_SI$Gender[is.na(dataMNAR_SI$Gender)] <- MNAR_SI$Gender
dataMNAR_SI$Age[is.na(dataMNAR_SI$Age)] <- MNAR_SI$Age
dataMNAR_SI$PerfEval[is.na(dataMNAR_SI$PerfEval)] <- MNAR_SI$PerfEval
dataMNAR_SI$Education[is.na(dataMNAR_SI$Education)] <- MNAR_SI$Education
dataMNAR_SI$Dept[is.na(dataMNAR_SI$Dept)] <- MNAR_SI$Dept
dataMNAR_SI$Seniority[is.na(dataMNAR_SI$Seniority)] <- MNAR_SI$Seniority
dataMNAR_SI$BasePay[is.na(dataMNAR_SI$BasePay)] <- MNAR_SI$BasePay
dataMNAR_SI$Bonus[is.na(dataMNAR_SI$Bonus)] <- MNAR_SI$Bonus
# Imputation using kNN
# Selection of k
k <- round(sqrt(nrow(data)))
k <- if_else(k %% 2 == 0, k+1,k)
# Impute MCAR NAs with kNN
dataMCAR_kNN <- dataMCAR
dataMCAR_kNN <- kNN(dataMCAR_kNN, k=k)
# Impute MAR NAs with kNN
dataMAR_kNN <- dataMAR
dataMAR_kNN <- kNN(dataMAR_kNN, k=k)
# Impute MNAR NAs with kNN
dataMNAR_kNN <- dataMNAR
dataMNAR_kNN <- kNN(dataMNAR_kNN, k=k)
# Imputation using Random Forest (RF)
# Impute MCAR NAs with RF
dataMCAR_RF <- dataMCAR
dataMCAR_RF$JobTitle <- as.factor(dataMCAR_RF$JobTitle)
dataMCAR_RF$Gender <- as.factor(dataMCAR_RF$Gender)
dataMCAR_RF$Education <- as.factor(dataMCAR_RF$Education)
dataMCAR_RF$Dept <- as.factor(dataMCAR_RF$Dept)
dataMCAR_RF <- missForest(dataMCAR_RF)
# Impute MAR NAs with RF
dataMAR_RF <- dataMAR
dataMAR_RF$JobTitle <- as.factor(dataMAR_RF$JobTitle)
dataMAR_RF$Gender <- as.factor(dataMAR_RF$Gender)
dataMAR_RF$Education <- as.factor(dataMAR_RF$Education)
dataMAR_RF$Dept <- as.factor(dataMAR_RF$Dept)
dataMAR_RF <- missForest(dataMAR_RF)
# Impute MNAR NAs with RF
dataMNAR_RF <- dataMNAR
dataMNAR_RF$JobTitle <- as.factor(dataMNAR_RF$JobTitle)
dataMNAR_RF$Gender <- as.factor(dataMNAR_RF$Gender)
dataMNAR_RF$Education <- as.factor(dataMNAR_RF$Education)
dataMNAR_RF$Dept <- as.factor(dataMNAR_RF$Dept)
dataMNAR_RF <- missForest(dataMNAR_RF)
# Imputation using MICE
# Impute MCAR NAs with MICE
dataMCAR_MICE <- dataMCAR
dataMCAR_MICE$JobTitle <- as.factor(dataMCAR_MICE$JobTitle)
dataMCAR_MICE$Gender <- as.factor(dataMCAR_MICE$Gender)
dataMCAR_MICE$Education <- as.factor(dataMCAR_MICE$Education)
dataMCAR_MICE$Dept <- as.factor(dataMCAR_MICE$Dept)
dataMCAR_MICE <- mice(dataMCAR_MICE, m=5, method=c("polyreg","logreg","pmm","polr","polr","polyreg","polr","pmm","pmm"), maxit=20)
dataMCAR_MICE_Final <- complete(dataMCAR_MICE,5)
# Impute MAR NAs with MICE
dataMAR_MICE <- dataMAR
dataMAR_MICE$JobTitle <- as.factor(dataMAR_MICE$JobTitle)
dataMAR_MICE$Gender <- as.factor(dataMAR_MICE$Gender)
dataMAR_MICE$Education <- as.factor(dataMAR_MICE$Education)
dataMAR_MICE$Dept <- as.factor(dataMAR_MICE$Dept)
dataMAR_MICE <- mice(dataMAR_MICE, m=5, method=c("polyreg","logreg","pmm","polr","polr","polyreg","polr","pmm","pmm"), maxit=20)
dataMAR_MICE_Final <- complete(dataMAR_MICE,5)
# Impute MNAR NAs with MICE
dataMNAR_MICE <- dataMNAR
dataMNAR_MICE$JobTitle <- as.factor(dataMNAR_MICE$JobTitle)
dataMNAR_MICE$Gender <- as.factor(dataMNAR_MICE$Gender)
dataMNAR_MICE$Education <- as.factor(dataMNAR_MICE$Education)
dataMNAR_MICE$Dept <- as.factor(dataMNAR_MICE$Dept)
dataMNAR_MICE <- mice(dataMNAR_MICE, m=5, method=c("polyreg","logreg","pmm","polr","polr","polyreg","polr","pmm","pmm"), maxit=20)
dataMNAR_MICE_Final <- complete(dataMNAR_MICE,5)
# Group Job Title Original and Imputed data
JobTitle_MCAR_Data <- tibble(JobTitle_Original = JobTitle_MCAR_NAs,
JobTitle_SI=dataMCAR_SI$JobTitle[filter(index_MCAR, index_MCAR$col==1)$row],
JobTitle_kNN=dataMCAR_kNN$JobTitle[filter(index_MCAR, index_MCAR$col==1)$row],
JobTitle_RF=dataMCAR_RF$ximp$JobTitle[filter(index_MCAR, index_MCAR$col==1)$row],
JobTitle_MICE=dataMCAR_MICE_Final$JobTitle[filter(index_MCAR, index_MCAR$col==1)$row])
JobTitle_MAR_Data <- tibble(JobTitle_Original = JobTitle_MAR_NAs,
JobTitle_SI=dataMAR_SI$JobTitle[filter(index_MAR, index_MAR$col==1)$row],
JobTitle_kNN=dataMAR_kNN$JobTitle[filter(index_MAR, index_MAR$col==1)$row],
JobTitle_RF=dataMAR_RF$ximp$JobTitle[filter(index_MAR, index_MAR$col==1)$row],
JobTitle_MICE=dataMAR_MICE_Final$JobTitle[filter(index_MAR, index_MAR$col==1)$row])
JobTitle_MNAR_Data <- tibble(JobTitle_Original = JobTitle_MNAR_NAs,
JobTitle_SI=dataMNAR_SI$JobTitle[filter(index_MNAR, index_MNAR$col==1)$row],
JobTitle_kNN=dataMNAR_kNN$JobTitle[filter(index_MNAR, index_MNAR$col==1)$row],
JobTitle_RF=dataMNAR_RF$ximp$JobTitle[filter(index_MNAR, index_MNAR$col==1)$row],
JobTitle_MICE=dataMNAR_MICE_Final$JobTitle[filter(index_MNAR, index_MNAR$col==1)$row])
# Add count variable for Job Title to compute error
# MCAR
JobTitle_MCAR_Data$JobTitle_SI_Count <- ifelse(JobTitle_MCAR_Data$JobTitle_Original==JobTitle_MCAR_Data$JobTitle_SI,as.integer(1),as.integer(0))
JobTitle_MCAR_Data$JobTitle_kNN_Count <- ifelse(JobTitle_MCAR_Data$JobTitle_Original==JobTitle_MCAR_Data$JobTitle_kNN,as.integer(1),as.integer(0))
JobTitle_MCAR_Data$JobTitle_RF_Count <- ifelse(JobTitle_MCAR_Data$JobTitle_Original==JobTitle_MCAR_Data$JobTitle_RF,as.integer(1),as.integer(0))
JobTitle_MCAR_Data$JobTitle_MICE_Count <- ifelse(JobTitle_MCAR_Data$JobTitle_Original==JobTitle_MCAR_Data$JobTitle_MICE,as.integer(1),as.integer(0))
# MAR
JobTitle_MAR_Data$JobTitle_SI_Count <- ifelse(JobTitle_MAR_Data$JobTitle_Original==JobTitle_MAR_Data$JobTitle_SI,as.integer(1),as.integer(0))
JobTitle_MAR_Data$JobTitle_kNN_Count <- ifelse(JobTitle_MAR_Data$JobTitle_Original==JobTitle_MAR_Data$JobTitle_kNN,as.integer(1),as.integer(0))
JobTitle_MAR_Data$JobTitle_RF_Count <- ifelse(JobTitle_MAR_Data$JobTitle_Original==JobTitle_MAR_Data$JobTitle_RF,as.integer(1),as.integer(0))
JobTitle_MAR_Data$JobTitle_MICE_Count <- ifelse(JobTitle_MAR_Data$JobTitle_Original==JobTitle_MAR_Data$JobTitle_MICE,as.integer(1),as.integer(0))
# MNAR
JobTitle_MNAR_Data$JobTitle_SI_Count <- ifelse(JobTitle_MNAR_Data$JobTitle_Original==JobTitle_MNAR_Data$JobTitle_SI,as.integer(1),as.integer(0))
JobTitle_MNAR_Data$JobTitle_kNN_Count <- ifelse(JobTitle_MNAR_Data$JobTitle_Original==JobTitle_MNAR_Data$JobTitle_kNN,as.integer(1),as.integer(0))
JobTitle_MNAR_Data$JobTitle_RF_Count <- ifelse(JobTitle_MNAR_Data$JobTitle_Original==JobTitle_MNAR_Data$JobTitle_RF,as.integer(1),as.integer(0))
JobTitle_MNAR_Data$JobTitle_MICE_Count <- ifelse(JobTitle_MNAR_Data$JobTitle_Original==JobTitle_MNAR_Data$JobTitle_MICE,as.integer(1),as.integer(0))
# Compute Percent of Error for Job Title
# MCAR Error Percentage
Theor_MCAR <- nrow(JobTitle_MCAR_Data)
JobTitle_MCAR_SI_Err <- abs(Theor_MCAR-length(which(JobTitle_MCAR_Data$JobTitle_SI_Count==1)))/Theor_MCAR*100
JobTitle_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(JobTitle_MCAR_Data$JobTitle_kNN_Count==1)))/Theor_MCAR*100
JobTitle_MCAR_RF_Err <- abs(Theor_MCAR-length(which(JobTitle_MCAR_Data$JobTitle_RF_Count==1)))/Theor_MCAR*100
JobTitle_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(JobTitle_MCAR_Data$JobTitle_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(JobTitle_MAR_Data)
JobTitle_MAR_SI_Err <- abs(Theor_MAR-length(which(JobTitle_MAR_Data$JobTitle_SI_Count==1)))/Theor_MAR*100
JobTitle_MAR_kNN_Err <- abs(Theor_MAR-length(which(JobTitle_MAR_Data$JobTitle_kNN_Count==1)))/Theor_MAR*100
JobTitle_MAR_RF_Err <- abs(Theor_MAR-length(which(JobTitle_MAR_Data$JobTitle_RF_Count==1)))/Theor_MAR*100
JobTitle_MAR_MICE_Err <- abs(Theor_MAR-length(which(JobTitle_MAR_Data$JobTitle_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(JobTitle_MNAR_Data)
JobTitle_MNAR_SI_Err <- abs(Theor_MNAR-length(which(JobTitle_MNAR_Data$JobTitle_SI_Count==1)))/Theor_MNAR*100
JobTitle_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(JobTitle_MNAR_Data$JobTitle_kNN_Count==1)))/Theor_MNAR*100
JobTitle_MNAR_RF_Err <- abs(Theor_MNAR-length(which(JobTitle_MNAR_Data$JobTitle_RF_Count==1)))/Theor_MNAR*100
JobTitle_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(JobTitle_MNAR_Data$JobTitle_MICE_Count==1)))/Theor_MNAR*100
JobTitle_Error <- data.frame(SI_Error = c(JobTitle_MCAR_SI_Err, JobTitle_MAR_SI_Err, JobTitle_MNAR_SI_Err),
kNN_Error = c(JobTitle_MCAR_kNN_Err, JobTitle_MAR_kNN_Err, JobTitle_MAR_kNN_Err),
RF_Error = c(JobTitle_MCAR_RF_Err, JobTitle_MAR_RF_Err, JobTitle_MAR_RF_Err),
MICE_Error = c(JobTitle_MCAR_MICE_Err, JobTitle_MAR_MICE_Err, JobTitle_MAR_MICE_Err))
rownames(JobTitle_Error) <- c("JobTitle_MCAR","JobTitle_MAR","JobTitle_MNAR")
# Group Gender Original and Imputed data
Gender_MCAR_Data <- tibble(Gender_Original = Gender_MCAR_NAs,
Gender_SI=dataMCAR_SI$Gender[filter(index_MCAR, index_MCAR$col==2)$row],
Gender_kNN=dataMCAR_kNN$Gender[filter(index_MCAR, index_MCAR$col==2)$row],
Gender_RF=dataMCAR_RF$ximp$Gender[filter(index_MCAR, index_MCAR$col==2)$row],
Gender_MICE=dataMCAR_MICE_Final$Gender[filter(index_MCAR, index_MCAR$col==2)$row])
Gender_MAR_Data <- tibble(Gender_Original = Gender_MAR_NAs,
Gender_SI=dataMAR_SI$Gender[filter(index_MAR, index_MAR$col==2)$row],
Gender_kNN=dataMAR_kNN$Gender[filter(index_MAR, index_MAR$col==2)$row],
Gender_RF=dataMAR_RF$ximp$Gender[filter(index_MAR, index_MAR$col==2)$row],
Gender_MICE=dataMAR_MICE_Final$Gender[filter(index_MAR, index_MAR$col==2)$row])
Gender_MNAR_Data <- tibble(Gender_Original = Gender_MNAR_NAs,
Gender_SI=dataMNAR_SI$Gender[filter(index_MNAR, index_MNAR$col==2)$row],
Gender_kNN=dataMNAR_kNN$Gender[filter(index_MNAR, index_MNAR$col==2)$row],
Gender_RF=dataMNAR_RF$ximp$Gender[filter(index_MNAR, index_MNAR$col==2)$row],
Gender_MICE=dataMNAR_MICE_Final$Gender[filter(index_MNAR, index_MNAR$col==2)$row])
# Add count variable for Job Title to compute error
# MCAR
Gender_MCAR_Data$Gender_SI_Count <- ifelse(Gender_MCAR_Data$Gender_Original==Gender_MCAR_Data$Gender_SI,as.integer(1),as.integer(0))
Gender_MCAR_Data$Gender_kNN_Count <- ifelse(Gender_MCAR_Data$Gender_Original==Gender_MCAR_Data$Gender_kNN,as.integer(1),as.integer(0))
Gender_MCAR_Data$Gender_RF_Count <- ifelse(Gender_MCAR_Data$Gender_Original==Gender_MCAR_Data$Gender_RF,as.integer(1),as.integer(0))
Gender_MCAR_Data$Gender_MICE_Count <- ifelse(Gender_MCAR_Data$Gender_Original==Gender_MCAR_Data$Gender_MICE,as.integer(1),as.integer(0))
# MAR
Gender_MAR_Data$Gender_SI_Count <- ifelse(Gender_MAR_Data$Gender_Original==Gender_MAR_Data$Gender_SI,as.integer(1),as.integer(0))
Gender_MAR_Data$Gender_kNN_Count <- ifelse(Gender_MAR_Data$Gender_Original==Gender_MAR_Data$Gender_kNN,as.integer(1),as.integer(0))
Gender_MAR_Data$Gender_RF_Count <- ifelse(Gender_MAR_Data$Gender_Original==Gender_MAR_Data$Gender_RF,as.integer(1),as.integer(0))
Gender_MAR_Data$Gender_MICE_Count <- ifelse(Gender_MAR_Data$Gender_Original==Gender_MAR_Data$Gender_MICE,as.integer(1),as.integer(0))
# MNAR
Gender_MNAR_Data$Gender_SI_Count <- ifelse(Gender_MNAR_Data$Gender_Original==Gender_MNAR_Data$Gender_SI,as.integer(1),as.integer(0))
Gender_MNAR_Data$Gender_kNN_Count <- ifelse(Gender_MNAR_Data$Gender_Original==Gender_MNAR_Data$Gender_kNN,as.integer(1),as.integer(0))
Gender_MNAR_Data$Gender_RF_Count <- ifelse(Gender_MNAR_Data$Gender_Original==Gender_MNAR_Data$Gender_RF,as.integer(1),as.integer(0))
Gender_MNAR_Data$Gender_MICE_Count <- ifelse(Gender_MNAR_Data$Gender_Original==Gender_MNAR_Data$Gender_MICE,as.integer(1),as.integer(0))
# Compute Percent of Error for Gender
# MCAR Error Percentage
Theor_MCAR <- nrow(Gender_MCAR_Data)
Gender_MCAR_SI_Err <- abs(Theor_MCAR-length(which(Gender_MCAR_Data$Gender_SI_Count==1)))/Theor_MCAR*100
Gender_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(Gender_MCAR_Data$Gender_kNN_Count==1)))/Theor_MCAR*100
Gender_MCAR_RF_Err <- abs(Theor_MCAR-length(which(Gender_MCAR_Data$Gender_RF_Count==1)))/Theor_MCAR*100
Gender_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(Gender_MCAR_Data$Gender_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(Gender_MAR_Data)
Gender_MAR_SI_Err <- abs(Theor_MAR-length(which(Gender_MAR_Data$Gender_SI_Count==1)))/Theor_MAR*100
Gender_MAR_kNN_Err <- abs(Theor_MAR-length(which(Gender_MAR_Data$Gender_kNN_Count==1)))/Theor_MAR*100
Gender_MAR_RF_Err <- abs(Theor_MAR-length(which(Gender_MAR_Data$Gender_RF_Count==1)))/Theor_MAR*100
Gender_MAR_MICE_Err <- abs(Theor_MAR-length(which(Gender_MAR_Data$Gender_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(Gender_MNAR_Data)
Gender_MNAR_SI_Err <- abs(Theor_MNAR-length(which(Gender_MNAR_Data$Gender_SI_Count==1)))/Theor_MNAR*100
Gender_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(Gender_MNAR_Data$Gender_kNN_Count==1)))/Theor_MNAR*100
Gender_MNAR_RF_Err <- abs(Theor_MNAR-length(which(Gender_MNAR_Data$Gender_RF_Count==1)))/Theor_MNAR*100
Gender_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(Gender_MNAR_Data$Gender_MICE_Count==1)))/Theor_MNAR*100
Gender_Error <- data.frame(SI_Error = c(Gender_MCAR_SI_Err, Gender_MAR_SI_Err, Gender_MNAR_SI_Err),
kNN_Error = c(Gender_MCAR_kNN_Err, Gender_MAR_kNN_Err, Gender_MAR_kNN_Err),
RF_Error = c(Gender_MCAR_RF_Err, Gender_MAR_RF_Err, Gender_MAR_RF_Err),
MICE_Error = c(Gender_MCAR_MICE_Err, Gender_MAR_MICE_Err, Gender_MAR_MICE_Err))
rownames(Gender_Error) <- c("Gender_MCAR","Gender_MAR","Gender_MNAR")
# Group Age Original and Imputed data
Age_MCAR_Data <- tibble(Age_Original = Age_MCAR_NAs,
Age_SI=dataMCAR_SI$Age[filter(index_MCAR, index_MCAR$col==3)$row],
Age_kNN=dataMCAR_kNN$Age[filter(index_MCAR, index_MCAR$col==3)$row],
Age_RF=dataMCAR_RF$ximp$Age[filter(index_MCAR, index_MCAR$col==3)$row],
Age_MICE=dataMCAR_MICE_Final$Age[filter(index_MCAR, index_MCAR$col==3)$row])
Age_MAR_Data <- tibble(Age_Original = Age_MAR_NAs,
Age_SI=dataMAR_SI$Age[filter(index_MAR, index_MAR$col==3)$row],
Age_kNN=dataMAR_kNN$Age[filter(index_MAR, index_MAR$col==3)$row],
Age_RF=dataMAR_RF$ximp$Age[filter(index_MAR, index_MAR$col==3)$row],
Age_MICE=dataMAR_MICE_Final$Age[filter(index_MAR, index_MAR$col==3)$row])
Age_MNAR_Data <- tibble(Age_Original = Age_MNAR_NAs,
Age_SI=dataMNAR_SI$Age[filter(index_MNAR, index_MNAR$col==3)$row],
Age_kNN=dataMNAR_kNN$Age[filter(index_MNAR, index_MNAR$col==3)$row],
Age_RF=dataMNAR_RF$ximp$Age[filter(index_MNAR, index_MNAR$col==3)$row],
Age_MICE=dataMNAR_MICE_Final$Age[filter(index_MNAR, index_MNAR$col==3)$row])
# Compute Percent of Error for Age
# MCAR Error Percentage
Age_MCAR_Error <- tibble(Age_SI_Error=abs(Age_MCAR_Data$Age_Original-Age_MCAR_Data$Age_SI)/Age_MCAR_Data$Age_Original*100,
Age_kNN_Error=abs(Age_MCAR_Data$Age_Original-Age_MCAR_Data$Age_kNN)/Age_MCAR_Data$Age_Original*100,
Age_RF_Error=abs(Age_MCAR_Data$Age_Original-Age_MCAR_Data$Age_RF)/Age_MCAR_Data$Age_Original*100,
Age_MICE_Error=abs(Age_MCAR_Data$Age_Original-Age_MCAR_Data$Age_MICE)/Age_MCAR_Data$Age_Original*100)
# MAR Error Percentage
Age_MAR_Error <- tibble(Age_SI_Error=abs(Age_MAR_Data$Age_Original-Age_MAR_Data$Age_SI)/Age_MAR_Data$Age_Original*100,
Age_kNN_Error=abs(Age_MAR_Data$Age_Original-Age_MAR_Data$Age_kNN)/Age_MAR_Data$Age_Original*100,
Age_RF_Error=abs(Age_MAR_Data$Age_Original-Age_MAR_Data$Age_RF)/Age_MAR_Data$Age_Original*100,
Age_MICE_Error=abs(Age_MAR_Data$Age_Original-Age_MAR_Data$Age_MICE)/Age_MAR_Data$Age_Original*100)
# MNAR Error Percentage
Age_MNAR_Error <- tibble(Age_SI_Error=abs(Age_MNAR_Data$Age_Original-Age_MNAR_Data$Age_SI)/Age_MNAR_Data$Age_Original*100,
Age_kNN_Error=abs(Age_MNAR_Data$Age_Original-Age_MNAR_Data$Age_kNN)/Age_MNAR_Data$Age_Original*100,
Age_RF_Error=abs(Age_MNAR_Data$Age_Original-Age_MNAR_Data$Age_RF)/Age_MNAR_Data$Age_Original*100,
Age_MICE_Error=abs(Age_MNAR_Data$Age_Original-Age_MNAR_Data$Age_MICE)/Age_MNAR_Data$Age_Original*100)
Age_Error <- data.frame(
SI_Error = c(mean(Age_MCAR_Error$Age_SI_Error) , mean(Age_MAR_Error$Age_SI_Error) , mean(Age_MNAR_Error$Age_SI_Error)),
kNN_Error = c(mean(Age_MCAR_Error$Age_kNN_Error) , mean(Age_MAR_Error$Age_kNN_Error) , mean(Age_MNAR_Error$Age_kNN_Error)),
RF_Error = c(mean(Age_MCAR_Error$Age_RF_Error) , mean(Age_MAR_Error$Age_RF_Error) , mean(Age_MNAR_Error$Age_RF_Error)),
MICE_Error = c(mean(Age_MCAR_Error$Age_MICE_Error), mean(Age_MAR_Error$Age_MICE_Error), mean(Age_MNAR_Error$Age_MICE_Error)))
rownames(Age_Error) <- c("Age_MCAR","Age_MAR","Age_MNAR")
# Group PerfEval Original and Imputed data
PerfEval_MCAR_Data <- tibble(PerfEval_Original = PerfEval_MCAR_NAs,
PerfEval_SI=dataMCAR_SI$PerfEval[filter(index_MCAR, index_MCAR$col==4)$row],
PerfEval_kNN=dataMCAR_kNN$PerfEval[filter(index_MCAR, index_MCAR$col==4)$row],
PerfEval_RF=dataMCAR_RF$ximp$PerfEval[filter(index_MCAR, index_MCAR$col==4)$row],
PerfEval_MICE=dataMCAR_MICE_Final$PerfEval[filter(index_MCAR, index_MCAR$col==4)$row])
PerfEval_MAR_Data <- tibble(PerfEval_Original = PerfEval_MAR_NAs,
PerfEval_SI=dataMAR_SI$PerfEval[filter(index_MAR, index_MAR$col==4)$row],
PerfEval_kNN=dataMAR_kNN$PerfEval[filter(index_MAR, index_MAR$col==4)$row],
PerfEval_RF=dataMAR_RF$ximp$PerfEval[filter(index_MAR, index_MAR$col==4)$row],
PerfEval_MICE=dataMAR_MICE_Final$PerfEval[filter(index_MAR, index_MAR$col==4)$row])
PerfEval_MNAR_Data <- tibble(PerfEval_Original = PerfEval_MNAR_NAs,
PerfEval_SI=dataMNAR_SI$PerfEval[filter(index_MNAR, index_MNAR$col==4)$row],
PerfEval_kNN=dataMNAR_kNN$PerfEval[filter(index_MNAR, index_MNAR$col==4)$row],
PerfEval_RF=dataMNAR_RF$ximp$PerfEval[filter(index_MNAR, index_MNAR$col==4)$row],
PerfEval_MICE=dataMNAR_MICE_Final$PerfEval[filter(index_MNAR, index_MNAR$col==4)$row])
# Add count variable for PerfEval to compute error
# MCAR
PerfEval_MCAR_Data$PerfEval_SI_Count <- ifelse(PerfEval_MCAR_Data$PerfEval_Original==PerfEval_MCAR_Data$PerfEval_SI,as.integer(1),as.integer(0))
PerfEval_MCAR_Data$PerfEval_kNN_Count <- ifelse(PerfEval_MCAR_Data$PerfEval_Original==PerfEval_MCAR_Data$PerfEval_kNN,as.integer(1),as.integer(0))
PerfEval_MCAR_Data$PerfEval_RF_Count <- ifelse(PerfEval_MCAR_Data$PerfEval_Original==PerfEval_MCAR_Data$PerfEval_RF,as.integer(1),as.integer(0))
PerfEval_MCAR_Data$PerfEval_MICE_Count <- ifelse(PerfEval_MCAR_Data$PerfEval_Original==PerfEval_MCAR_Data$PerfEval_MICE,as.integer(1),as.integer(0))
# MAR
PerfEval_MAR_Data$PerfEval_SI_Count <- ifelse(PerfEval_MAR_Data$PerfEval_Original==PerfEval_MAR_Data$PerfEval_SI,as.integer(1),as.integer(0))
PerfEval_MAR_Data$PerfEval_kNN_Count <- ifelse(PerfEval_MAR_Data$PerfEval_Original==PerfEval_MAR_Data$PerfEval_kNN,as.integer(1),as.integer(0))
PerfEval_MAR_Data$PerfEval_RF_Count <- ifelse(PerfEval_MAR_Data$PerfEval_Original==PerfEval_MAR_Data$PerfEval_RF,as.integer(1),as.integer(0))
PerfEval_MAR_Data$PerfEval_MICE_Count <- ifelse(PerfEval_MAR_Data$PerfEval_Original==PerfEval_MAR_Data$PerfEval_MICE,as.integer(1),as.integer(0))
# MNAR
PerfEval_MNAR_Data$PerfEval_SI_Count <- ifelse(PerfEval_MNAR_Data$PerfEval_Original==PerfEval_MNAR_Data$PerfEval_SI,as.integer(1),as.integer(0))
PerfEval_MNAR_Data$PerfEval_kNN_Count <- ifelse(PerfEval_MNAR_Data$PerfEval_Original==PerfEval_MNAR_Data$PerfEval_kNN,as.integer(1),as.integer(0))
PerfEval_MNAR_Data$PerfEval_RF_Count <- ifelse(PerfEval_MNAR_Data$PerfEval_Original==PerfEval_MNAR_Data$PerfEval_RF,as.integer(1),as.integer(0))
PerfEval_MNAR_Data$PerfEval_MICE_Count <- ifelse(PerfEval_MNAR_Data$PerfEval_Original==PerfEval_MNAR_Data$PerfEval_MICE,as.integer(1),as.integer(0))
# Compute Percent of Error for PerfEval
# MCAR Error Percentage
Theor_MCAR <- nrow(PerfEval_MCAR_Data)
PerfEval_MCAR_SI_Err <- abs(Theor_MCAR-length(which(PerfEval_MCAR_Data$PerfEval_SI_Count==1)))/Theor_MCAR*100
PerfEval_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(PerfEval_MCAR_Data$PerfEval_kNN_Count==1)))/Theor_MCAR*100
PerfEval_MCAR_RF_Err <- abs(Theor_MCAR-length(which(PerfEval_MCAR_Data$PerfEval_RF_Count==1)))/Theor_MCAR*100
PerfEval_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(PerfEval_MCAR_Data$PerfEval_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(PerfEval_MAR_Data)
PerfEval_MAR_SI_Err <- abs(Theor_MAR-length(which(PerfEval_MAR_Data$PerfEval_SI_Count==1)))/Theor_MAR*100
PerfEval_MAR_kNN_Err <- abs(Theor_MAR-length(which(PerfEval_MAR_Data$PerfEval_kNN_Count==1)))/Theor_MAR*100
PerfEval_MAR_RF_Err <- abs(Theor_MAR-length(which(PerfEval_MAR_Data$PerfEval_RF_Count==1)))/Theor_MAR*100
PerfEval_MAR_MICE_Err <- abs(Theor_MAR-length(which(PerfEval_MAR_Data$PerfEval_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(PerfEval_MNAR_Data)
PerfEval_MNAR_SI_Err <- abs(Theor_MNAR-length(which(PerfEval_MNAR_Data$PerfEval_SI_Count==1)))/Theor_MNAR*100
PerfEval_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(PerfEval_MNAR_Data$PerfEval_kNN_Count==1)))/Theor_MNAR*100
PerfEval_MNAR_RF_Err <- abs(Theor_MNAR-length(which(PerfEval_MNAR_Data$PerfEval_RF_Count==1)))/Theor_MNAR*100
PerfEval_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(PerfEval_MNAR_Data$PerfEval_MICE_Count==1)))/Theor_MNAR*100
PerfEval_Error <- data.frame(SI_Error = c(PerfEval_MCAR_SI_Err, PerfEval_MAR_SI_Err, PerfEval_MNAR_SI_Err),
kNN_Error = c(PerfEval_MCAR_kNN_Err, PerfEval_MAR_kNN_Err, PerfEval_MAR_kNN_Err),
RF_Error = c(PerfEval_MCAR_RF_Err, PerfEval_MAR_RF_Err, PerfEval_MAR_RF_Err),
MICE_Error = c(PerfEval_MCAR_MICE_Err, PerfEval_MAR_MICE_Err, PerfEval_MAR_MICE_Err))
rownames(PerfEval_Error) <- c("PerfEval_MCAR","PerfEval_MAR","PerfEval_MNAR")
# Group Education Original and Imputed data
Education_MCAR_Data <- tibble(Education_Original = Education_MCAR_NAs,
Education_SI=dataMCAR_SI$Education[filter(index_MCAR, index_MCAR$col==5)$row],
Education_kNN=dataMCAR_kNN$Education[filter(index_MCAR, index_MCAR$col==5)$row],
Education_RF=dataMCAR_RF$ximp$Education[filter(index_MCAR, index_MCAR$col==5)$row],
Education_MICE=dataMCAR_MICE_Final$Education[filter(index_MCAR, index_MCAR$col==5)$row])
Education_MAR_Data <- tibble(Education_Original = Education_MAR_NAs,
Education_SI=dataMAR_SI$Education[filter(index_MAR, index_MAR$col==5)$row],
Education_kNN=dataMAR_kNN$Education[filter(index_MAR, index_MAR$col==5)$row],
Education_RF=dataMAR_RF$ximp$Education[filter(index_MAR, index_MAR$col==5)$row],
Education_MICE=dataMAR_MICE_Final$Education[filter(index_MAR, index_MAR$col==5)$row])
Education_MNAR_Data <- tibble(Education_Original = Education_MNAR_NAs,
Education_SI=dataMNAR_SI$Education[filter(index_MNAR, index_MNAR$col==5)$row],
Education_kNN=dataMNAR_kNN$Education[filter(index_MNAR, index_MNAR$col==5)$row],
Education_RF=dataMNAR_RF$ximp$Education[filter(index_MNAR, index_MNAR$col==5)$row],
Education_MICE=dataMNAR_MICE_Final$Education[filter(index_MNAR, index_MNAR$col==5)$row])
# Add count variable for Education to compute error
# MCAR
Education_MCAR_Data$Education_SI_Count <- ifelse(Education_MCAR_Data$Education_Original==Education_MCAR_Data$Education_SI,as.integer(1),as.integer(0))
Education_MCAR_Data$Education_kNN_Count <- ifelse(Education_MCAR_Data$Education_Original==Education_MCAR_Data$Education_kNN,as.integer(1),as.integer(0))
Education_MCAR_Data$Education_RF_Count <- ifelse(Education_MCAR_Data$Education_Original==Education_MCAR_Data$Education_RF,as.integer(1),as.integer(0))
Education_MCAR_Data$Education_MICE_Count <- ifelse(Education_MCAR_Data$Education_Original==Education_MCAR_Data$Education_MICE,as.integer(1),as.integer(0))
# MAR
Education_MAR_Data$Education_SI_Count <- ifelse(Education_MAR_Data$Education_Original==Education_MAR_Data$Education_SI,as.integer(1),as.integer(0))
Education_MAR_Data$Education_kNN_Count <- ifelse(Education_MAR_Data$Education_Original==Education_MAR_Data$Education_kNN,as.integer(1),as.integer(0))
Education_MAR_Data$Education_RF_Count <- ifelse(Education_MAR_Data$Education_Original==Education_MAR_Data$Education_RF,as.integer(1),as.integer(0))
Education_MAR_Data$Education_MICE_Count <- ifelse(Education_MAR_Data$Education_Original==Education_MAR_Data$Education_MICE,as.integer(1),as.integer(0))
# MNAR
Education_MNAR_Data$Education_SI_Count <- ifelse(Education_MNAR_Data$Education_Original==Education_MNAR_Data$Education_SI,as.integer(1),as.integer(0))
Education_MNAR_Data$Education_kNN_Count <- ifelse(Education_MNAR_Data$Education_Original==Education_MNAR_Data$Education_kNN,as.integer(1),as.integer(0))
Education_MNAR_Data$Education_RF_Count <- ifelse(Education_MNAR_Data$Education_Original==Education_MNAR_Data$Education_RF,as.integer(1),as.integer(0))
Education_MNAR_Data$Education_MICE_Count <- ifelse(Education_MNAR_Data$Education_Original==Education_MNAR_Data$Education_MICE,as.integer(1),as.integer(0))
# Compute Percent of Error for Education
# MCAR Error Percentage
Theor_MCAR <- nrow(Education_MCAR_Data)
Education_MCAR_SI_Err <- abs(Theor_MCAR-length(which(Education_MCAR_Data$Education_SI_Count==1)))/Theor_MCAR*100
Education_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(Education_MCAR_Data$Education_kNN_Count==1)))/Theor_MCAR*100
Education_MCAR_RF_Err <- abs(Theor_MCAR-length(which(Education_MCAR_Data$Education_RF_Count==1)))/Theor_MCAR*100
Education_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(Education_MCAR_Data$Education_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(Education_MAR_Data)
Education_MAR_SI_Err <- abs(Theor_MAR-length(which(Education_MAR_Data$Education_SI_Count==1)))/Theor_MAR*100
Education_MAR_kNN_Err <- abs(Theor_MAR-length(which(Education_MAR_Data$Education_kNN_Count==1)))/Theor_MAR*100
Education_MAR_RF_Err <- abs(Theor_MAR-length(which(Education_MAR_Data$Education_RF_Count==1)))/Theor_MAR*100
Education_MAR_MICE_Err <- abs(Theor_MAR-length(which(Education_MAR_Data$Education_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(Education_MNAR_Data)
Education_MNAR_SI_Err <- abs(Theor_MNAR-length(which(Education_MNAR_Data$Education_SI_Count==1)))/Theor_MNAR*100
Education_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(Education_MNAR_Data$Education_kNN_Count==1)))/Theor_MNAR*100
Education_MNAR_RF_Err <- abs(Theor_MNAR-length(which(Education_MNAR_Data$Education_RF_Count==1)))/Theor_MNAR*100
Education_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(Education_MNAR_Data$Education_MICE_Count==1)))/Theor_MNAR*100
Education_Error <- data.frame(SI_Error = c(Education_MCAR_SI_Err, Education_MAR_SI_Err, Education_MNAR_SI_Err),
kNN_Error = c(Education_MCAR_kNN_Err, Education_MAR_kNN_Err, Education_MAR_kNN_Err),
RF_Error = c(Education_MCAR_RF_Err, Education_MAR_RF_Err, Education_MAR_RF_Err),
MICE_Error = c(Education_MCAR_MICE_Err, Education_MAR_MICE_Err, Education_MAR_MICE_Err))
rownames(Education_Error) <- c("Education_MCAR","Education_MAR","Education_MNAR")
# Group Dept Original and Imputed data
Dept_MCAR_Data <- tibble(Dept_Original = Dept_MCAR_NAs,
Dept_SI=dataMCAR_SI$Dept[filter(index_MCAR, index_MCAR$col==6)$row],
Dept_kNN=dataMCAR_kNN$Dept[filter(index_MCAR, index_MCAR$col==6)$row],
Dept_RF=dataMCAR_RF$ximp$Dept[filter(index_MCAR, index_MCAR$col==6)$row],
Dept_MICE=dataMCAR_MICE_Final$Dept[filter(index_MCAR, index_MCAR$col==6)$row])
Dept_MAR_Data <- tibble(Dept_Original = Dept_MAR_NAs,
Dept_SI=dataMAR_SI$Dept[filter(index_MAR, index_MAR$col==6)$row],
Dept_kNN=dataMAR_kNN$Dept[filter(index_MAR, index_MAR$col==6)$row],
Dept_RF=dataMAR_RF$ximp$Dept[filter(index_MAR, index_MAR$col==6)$row],
Dept_MICE=dataMAR_MICE_Final$Dept[filter(index_MAR, index_MAR$col==6)$row])
Dept_MNAR_Data <- tibble(Dept_Original = Dept_MNAR_NAs,
Dept_SI=dataMNAR_SI$Dept[filter(index_MNAR, index_MNAR$col==6)$row],
Dept_kNN=dataMNAR_kNN$Dept[filter(index_MNAR, index_MNAR$col==6)$row],
Dept_RF=dataMNAR_RF$ximp$Dept[filter(index_MNAR, index_MNAR$col==6)$row],
Dept_MICE=dataMNAR_MICE_Final$Dept[filter(index_MNAR, index_MNAR$col==6)$row])
# Add count variable for Dept to compute error
# MCAR
Dept_MCAR_Data$Dept_SI_Count <- ifelse(Dept_MCAR_Data$Dept_Original==Dept_MCAR_Data$Dept_SI,as.integer(1),as.integer(0))
Dept_MCAR_Data$Dept_kNN_Count <- ifelse(Dept_MCAR_Data$Dept_Original==Dept_MCAR_Data$Dept_kNN,as.integer(1),as.integer(0))
Dept_MCAR_Data$Dept_RF_Count <- ifelse(Dept_MCAR_Data$Dept_Original==Dept_MCAR_Data$Dept_RF,as.integer(1),as.integer(0))
Dept_MCAR_Data$Dept_MICE_Count <- ifelse(Dept_MCAR_Data$Dept_Original==Dept_MCAR_Data$Dept_MICE,as.integer(1),as.integer(0))
# MAR
Dept_MAR_Data$Dept_SI_Count <- ifelse(Dept_MAR_Data$Dept_Original==Dept_MAR_Data$Dept_SI,as.integer(1),as.integer(0))
Dept_MAR_Data$Dept_kNN_Count <- ifelse(Dept_MAR_Data$Dept_Original==Dept_MAR_Data$Dept_kNN,as.integer(1),as.integer(0))
Dept_MAR_Data$Dept_RF_Count <- ifelse(Dept_MAR_Data$Dept_Original==Dept_MAR_Data$Dept_RF,as.integer(1),as.integer(0))
Dept_MAR_Data$Dept_MICE_Count <- ifelse(Dept_MAR_Data$Dept_Original==Dept_MAR_Data$Dept_MICE,as.integer(1),as.integer(0))
# MNAR
Dept_MNAR_Data$Dept_SI_Count <- ifelse(Dept_MNAR_Data$Dept_Original==Dept_MNAR_Data$Dept_SI,as.integer(1),as.integer(0))
Dept_MNAR_Data$Dept_kNN_Count <- ifelse(Dept_MNAR_Data$Dept_Original==Dept_MNAR_Data$Dept_kNN,as.integer(1),as.integer(0))
Dept_MNAR_Data$Dept_RF_Count <- ifelse(Dept_MNAR_Data$Dept_Original==Dept_MNAR_Data$Dept_RF,as.integer(1),as.integer(0))
Dept_MNAR_Data$Dept_MICE_Count <- ifelse(Dept_MNAR_Data$Dept_Original==Dept_MNAR_Data$Dept_MICE,as.integer(1),as.integer(0))
# Compute Percent of Error for Dept
# MCAR Error Percentage
Theor_MCAR <- nrow(Dept_MCAR_Data)
Dept_MCAR_SI_Err <- abs(Theor_MCAR-length(which(Dept_MCAR_Data$Dept_SI_Count==1)))/Theor_MCAR*100
Dept_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(Dept_MCAR_Data$Dept_kNN_Count==1)))/Theor_MCAR*100
Dept_MCAR_RF_Err <- abs(Theor_MCAR-length(which(Dept_MCAR_Data$Dept_RF_Count==1)))/Theor_MCAR*100
Dept_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(Dept_MCAR_Data$Dept_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(Dept_MAR_Data)
Dept_MAR_SI_Err <- abs(Theor_MAR-length(which(Dept_MAR_Data$Dept_SI_Count==1)))/Theor_MAR*100
Dept_MAR_kNN_Err <- abs(Theor_MAR-length(which(Dept_MAR_Data$Dept_kNN_Count==1)))/Theor_MAR*100
Dept_MAR_RF_Err <- abs(Theor_MAR-length(which(Dept_MAR_Data$Dept_RF_Count==1)))/Theor_MAR*100
Dept_MAR_MICE_Err <- abs(Theor_MAR-length(which(Dept_MAR_Data$Dept_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(Dept_MNAR_Data)
Dept_MNAR_SI_Err <- abs(Theor_MNAR-length(which(Dept_MNAR_Data$Dept_SI_Count==1)))/Theor_MNAR*100
Dept_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(Dept_MNAR_Data$Dept_kNN_Count==1)))/Theor_MNAR*100
Dept_MNAR_RF_Err <- abs(Theor_MNAR-length(which(Dept_MNAR_Data$Dept_RF_Count==1)))/Theor_MNAR*100
Dept_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(Dept_MNAR_Data$Dept_MICE_Count==1)))/Theor_MNAR*100
Dept_Error <- data.frame(SI_Error = c(Dept_MCAR_SI_Err, Dept_MAR_SI_Err, Dept_MNAR_SI_Err),
kNN_Error = c(Dept_MCAR_kNN_Err, Dept_MAR_kNN_Err, Dept_MAR_kNN_Err),
RF_Error = c(Dept_MCAR_RF_Err, Dept_MAR_RF_Err, Dept_MAR_RF_Err),
MICE_Error = c(Dept_MCAR_MICE_Err, Dept_MAR_MICE_Err, Dept_MAR_MICE_Err))
rownames(Dept_Error) <- c("Dept_MCAR","Dept_MAR","Dept_MNAR")
# Group Seniority Original and Imputed data
Seniority_MCAR_Data <- tibble(Seniority_Original = Seniority_MCAR_NAs,
Seniority_SI=dataMCAR_SI$Seniority[filter(index_MCAR, index_MCAR$col==7)$row],
Seniority_kNN=dataMCAR_kNN$Seniority[filter(index_MCAR, index_MCAR$col==7)$row],
Seniority_RF=dataMCAR_RF$ximp$Seniority[filter(index_MCAR, index_MCAR$col==7)$row],
Seniority_MICE=dataMCAR_MICE_Final$Seniority[filter(index_MCAR, index_MCAR$col==7)$row])
Seniority_MAR_Data <- tibble(Seniority_Original = Seniority_MAR_NAs,
Seniority_SI=dataMAR_SI$Seniority[filter(index_MAR, index_MAR$col==7)$row],
Seniority_kNN=dataMAR_kNN$Seniority[filter(index_MAR, index_MAR$col==7)$row],
Seniority_RF=dataMAR_RF$ximp$Seniority[filter(index_MAR, index_MAR$col==7)$row],
Seniority_MICE=dataMAR_MICE_Final$Seniority[filter(index_MAR, index_MAR$col==7)$row])
Seniority_MNAR_Data <- tibble(Seniority_Original = Seniority_MNAR_NAs,
Seniority_SI=dataMNAR_SI$Seniority[filter(index_MNAR, index_MNAR$col==7)$row],
Seniority_kNN=dataMNAR_kNN$Seniority[filter(index_MNAR, index_MNAR$col==7)$row],
Seniority_RF=dataMNAR_RF$ximp$Seniority[filter(index_MNAR, index_MNAR$col==7)$row],
Seniority_MICE=dataMNAR_MICE_Final$Seniority[filter(index_MNAR, index_MNAR$col==7)$row])
# Add count variable for Seniority to compute error
# MCAR
Seniority_MCAR_Data$Seniority_SI_Count <- ifelse(Seniority_MCAR_Data$Seniority_Original==Seniority_MCAR_Data$Seniority_SI,as.integer(1),as.integer(0))
Seniority_MCAR_Data$Seniority_kNN_Count <- ifelse(Seniority_MCAR_Data$Seniority_Original==Seniority_MCAR_Data$Seniority_kNN,as.integer(1),as.integer(0))
Seniority_MCAR_Data$Seniority_RF_Count <- ifelse(Seniority_MCAR_Data$Seniority_Original==Seniority_MCAR_Data$Seniority_RF,as.integer(1),as.integer(0))
Seniority_MCAR_Data$Seniority_MICE_Count <- ifelse(Seniority_MCAR_Data$Seniority_Original==Seniority_MCAR_Data$Seniority_MICE,as.integer(1),as.integer(0))
# MAR
Seniority_MAR_Data$Seniority_SI_Count <- ifelse(Seniority_MAR_Data$Seniority_Original==Seniority_MAR_Data$Seniority_SI,as.integer(1),as.integer(0))
Seniority_MAR_Data$Seniority_kNN_Count <- ifelse(Seniority_MAR_Data$Seniority_Original==Seniority_MAR_Data$Seniority_kNN,as.integer(1),as.integer(0))
Seniority_MAR_Data$Seniority_RF_Count <- ifelse(Seniority_MAR_Data$Seniority_Original==Seniority_MAR_Data$Seniority_RF,as.integer(1),as.integer(0))
Seniority_MAR_Data$Seniority_MICE_Count <- ifelse(Seniority_MAR_Data$Seniority_Original==Seniority_MAR_Data$Seniority_MICE,as.integer(1),as.integer(0))
# MNAR
Seniority_MNAR_Data$Seniority_SI_Count <- ifelse(Seniority_MNAR_Data$Seniority_Original==Seniority_MNAR_Data$Seniority_SI,as.integer(1),as.integer(0))
Seniority_MNAR_Data$Seniority_kNN_Count <- ifelse(Seniority_MNAR_Data$Seniority_Original==Seniority_MNAR_Data$Seniority_kNN,as.integer(1),as.integer(0))
Seniority_MNAR_Data$Seniority_RF_Count <- ifelse(Seniority_MNAR_Data$Seniority_Original==Seniority_MNAR_Data$Seniority_RF,as.integer(1),as.integer(0))
Seniority_MNAR_Data$Seniority_MICE_Count <- ifelse(Seniority_MNAR_Data$Seniority_Original==Seniority_MNAR_Data$Seniority_MICE,as.integer(1),as.integer(0))
# Seniority Error
# MCAR Error Percentage
Theor_MCAR <- nrow(Seniority_MCAR_Data)
Seniority_MCAR_SI_Err <- abs(Theor_MCAR-length(which(Seniority_MCAR_Data$Seniority_SI_Count==1)))/Theor_MCAR*100
Seniority_MCAR_kNN_Err <- abs(Theor_MCAR-length(which(Seniority_MCAR_Data$Seniority_kNN_Count==1)))/Theor_MCAR*100
Seniority_MCAR_RF_Err <- abs(Theor_MCAR-length(which(Seniority_MCAR_Data$Seniority_RF_Count==1)))/Theor_MCAR*100
Seniority_MCAR_MICE_Err <- abs(Theor_MCAR-length(which(Seniority_MCAR_Data$Seniority_MICE_Count==1)))/Theor_MCAR*100
# MAR Error Percentage
Theor_MAR <- nrow(Seniority_MAR_Data)
Seniority_MAR_SI_Err <- abs(Theor_MAR-length(which(Seniority_MAR_Data$Seniority_SI_Count==1)))/Theor_MAR*100
Seniority_MAR_kNN_Err <- abs(Theor_MAR-length(which(Seniority_MAR_Data$Seniority_kNN_Count==1)))/Theor_MAR*100
Seniority_MAR_RF_Err <- abs(Theor_MAR-length(which(Seniority_MAR_Data$Seniority_RF_Count==1)))/Theor_MAR*100
Seniority_MAR_MICE_Err <- abs(Theor_MAR-length(which(Seniority_MAR_Data$Seniority_MICE_Count==1)))/Theor_MAR*100
# MNAR Error Percentage
Theor_MNAR <- nrow(Seniority_MNAR_Data)
Seniority_MNAR_SI_Err <- abs(Theor_MNAR-length(which(Seniority_MNAR_Data$Seniority_SI_Count==1)))/Theor_MNAR*100
Seniority_MNAR_kNN_Err <- abs(Theor_MNAR-length(which(Seniority_MNAR_Data$Seniority_kNN_Count==1)))/Theor_MNAR*100
Seniority_MNAR_RF_Err <- abs(Theor_MNAR-length(which(Seniority_MNAR_Data$Seniority_RF_Count==1)))/Theor_MNAR*100
Seniority_MNAR_MICE_Err <- abs(Theor_MNAR-length(which(Seniority_MNAR_Data$Seniority_MICE_Count==1)))/Theor_MNAR*100
Seniority_Error <- data.frame(SI_Error = c(Seniority_MCAR_SI_Err, Seniority_MAR_SI_Err, Seniority_MNAR_SI_Err),
kNN_Error = c(Seniority_MCAR_kNN_Err, Seniority_MAR_kNN_Err, Seniority_MAR_kNN_Err),
RF_Error = c(Seniority_MCAR_RF_Err, Seniority_MAR_RF_Err, Seniority_MAR_RF_Err),
MICE_Error = c(Seniority_MCAR_MICE_Err, Seniority_MAR_MICE_Err, Seniority_MAR_MICE_Err))
rownames(Seniority_Error) <- c("Seniority_MCAR","Seniority_MAR","Seniority_MNAR")
# Group BasePay Original and Imputed data
BasePay_MCAR_Data <- tibble(BasePay_Original = BasePay_MCAR_NAs,
BasePay_SI=dataMCAR_SI$BasePay[filter(index_MCAR, index_MCAR$col==8)$row],
BasePay_kNN=dataMCAR_kNN$BasePay[filter(index_MCAR, index_MCAR$col==8)$row],
BasePay_RF=dataMCAR_RF$ximp$BasePay[filter(index_MCAR, index_MCAR$col==8)$row],
BasePay_MICE=dataMCAR_MICE_Final$BasePay[filter(index_MCAR, index_MCAR$col==8)$row])
BasePay_MAR_Data <- tibble(BasePay_Original = BasePay_MAR_NAs,
BasePay_SI=dataMAR_SI$BasePay[filter(index_MAR, index_MAR$col==8)$row],
BasePay_kNN=dataMAR_kNN$BasePay[filter(index_MAR, index_MAR$col==8)$row],
BasePay_RF=dataMAR_RF$ximp$BasePay[filter(index_MAR, index_MAR$col==8)$row],
BasePay_MICE=dataMAR_MICE_Final$BasePay[filter(index_MAR, index_MAR$col==8)$row])
BasePay_MNAR_Data <- tibble(BasePay_Original = BasePay_MNAR_NAs,
BasePay_SI=dataMNAR_SI$BasePay[filter(index_MNAR, index_MNAR$col==8)$row],
BasePay_kNN=dataMNAR_kNN$BasePay[filter(index_MNAR, index_MNAR$col==8)$row],
BasePay_RF=dataMNAR_RF$ximp$BasePay[filter(index_MNAR, index_MNAR$col==8)$row],
BasePay_MICE=dataMNAR_MICE_Final$BasePay[filter(index_MNAR, index_MNAR$col==8)$row])
# Compute Percent of Error for BasePay
# MCAR Error PercentBasePay
BasePay_MCAR_Error <- tibble(BasePay_SI_Error=abs(BasePay_MCAR_Data$BasePay_Original-BasePay_MCAR_Data$BasePay_SI)/BasePay_MCAR_Data$BasePay_Original*100,
BasePay_kNN_Error=abs(BasePay_MCAR_Data$BasePay_Original-BasePay_MCAR_Data$BasePay_kNN)/BasePay_MCAR_Data$BasePay_Original*100,
BasePay_RF_Error=abs(BasePay_MCAR_Data$BasePay_Original-BasePay_MCAR_Data$BasePay_RF)/BasePay_MCAR_Data$BasePay_Original*100,
BasePay_MICE_Error=abs(BasePay_MCAR_Data$BasePay_Original-BasePay_MCAR_Data$BasePay_MICE)/BasePay_MCAR_Data$BasePay_Original*100)
# MAR Error PercentBasePay
BasePay_MAR_Error <- tibble(BasePay_SI_Error=abs(BasePay_MAR_Data$BasePay_Original-BasePay_MAR_Data$BasePay_SI)/BasePay_MAR_Data$BasePay_Original*100,
BasePay_kNN_Error=abs(BasePay_MAR_Data$BasePay_Original-BasePay_MAR_Data$BasePay_kNN)/BasePay_MAR_Data$BasePay_Original*100,
BasePay_RF_Error=abs(BasePay_MAR_Data$BasePay_Original-BasePay_MAR_Data$BasePay_RF)/BasePay_MAR_Data$BasePay_Original*100,
BasePay_MICE_Error=abs(BasePay_MAR_Data$BasePay_Original-BasePay_MAR_Data$BasePay_MICE)/BasePay_MAR_Data$BasePay_Original*100)
# MNAR Error PercentBasePay
BasePay_MNAR_Error <- tibble(BasePay_SI_Error=abs(BasePay_MNAR_Data$BasePay_Original-BasePay_MNAR_Data$BasePay_SI)/BasePay_MNAR_Data$BasePay_Original*100,
BasePay_kNN_Error=abs(BasePay_MNAR_Data$BasePay_Original-BasePay_MNAR_Data$BasePay_kNN)/BasePay_MNAR_Data$BasePay_Original*100,
BasePay_RF_Error=abs(BasePay_MNAR_Data$BasePay_Original-BasePay_MNAR_Data$BasePay_RF)/BasePay_MNAR_Data$BasePay_Original*100,
BasePay_MICE_Error=abs(BasePay_MNAR_Data$BasePay_Original-BasePay_MNAR_Data$BasePay_MICE)/BasePay_MNAR_Data$BasePay_Original*100)
BasePay_Error <- data.frame(
SI_Error = c(mean(BasePay_MCAR_Error$BasePay_SI_Error) , mean(BasePay_MAR_Error$BasePay_SI_Error) , mean(BasePay_MNAR_Error$BasePay_SI_Error)),
kNN_Error = c(mean(BasePay_MCAR_Error$BasePay_kNN_Error) , mean(BasePay_MAR_Error$BasePay_kNN_Error) , mean(BasePay_MNAR_Error$BasePay_kNN_Error)),
RF_Error = c(mean(BasePay_MCAR_Error$BasePay_RF_Error) , mean(BasePay_MAR_Error$BasePay_RF_Error) , mean(BasePay_MNAR_Error$BasePay_RF_Error)),
MICE_Error = c(mean(BasePay_MCAR_Error$BasePay_MICE_Error), mean(BasePay_MAR_Error$BasePay_MICE_Error), mean(BasePay_MNAR_Error$BasePay_MICE_Error)))
rownames(BasePay_Error) <- c("BasePay_MCAR","BasePay_MAR","BasePay_MNAR")
# Group Bonus Original and Imputed data
Bonus_MCAR_Data <- tibble(Bonus_Original = Bonus_MCAR_NAs,
Bonus_SI=dataMCAR_SI$Bonus[filter(index_MCAR, index_MCAR$col==9)$row],
Bonus_kNN=dataMCAR_kNN$Bonus[filter(index_MCAR, index_MCAR$col==9)$row],
Bonus_RF=dataMCAR_RF$ximp$Bonus[filter(index_MCAR, index_MCAR$col==9)$row],
Bonus_MICE=dataMCAR_MICE_Final$Bonus[filter(index_MCAR, index_MCAR$col==9)$row])
Bonus_MAR_Data <- tibble(Bonus_Original = Bonus_MAR_NAs,
Bonus_SI=dataMAR_SI$Bonus[filter(index_MAR, index_MAR$col==9)$row],
Bonus_kNN=dataMAR_kNN$Bonus[filter(index_MAR, index_MAR$col==9)$row],
Bonus_RF=dataMAR_RF$ximp$Bonus[filter(index_MAR, index_MAR$col==9)$row],
Bonus_MICE=dataMAR_MICE_Final$Bonus[filter(index_MAR, index_MAR$col==9)$row])
Bonus_MNAR_Data <- tibble(Bonus_Original = Bonus_MNAR_NAs,
Bonus_SI=dataMNAR_SI$Bonus[filter(index_MNAR, index_MNAR$col==9)$row],
Bonus_kNN=dataMNAR_kNN$Bonus[filter(index_MNAR, index_MNAR$col==9)$row],
Bonus_RF=dataMNAR_RF$ximp$Bonus[filter(index_MNAR, index_MNAR$col==9)$row],
Bonus_MICE=dataMNAR_MICE_Final$Bonus[filter(index_MNAR, index_MNAR$col==9)$row])
# Compute Percent of Error for Bonus
# MCAR Error PercentBonus
Bonus_MCAR_Error <- tibble(Bonus_SI_Error=abs(Bonus_MCAR_Data$Bonus_Original-Bonus_MCAR_Data$Bonus_SI)/Bonus_MCAR_Data$Bonus_Original*100,
Bonus_kNN_Error=abs(Bonus_MCAR_Data$Bonus_Original-Bonus_MCAR_Data$Bonus_kNN)/Bonus_MCAR_Data$Bonus_Original*100,
Bonus_RF_Error=abs(Bonus_MCAR_Data$Bonus_Original-Bonus_MCAR_Data$Bonus_RF)/Bonus_MCAR_Data$Bonus_Original*100,
Bonus_MICE_Error=abs(Bonus_MCAR_Data$Bonus_Original-Bonus_MCAR_Data$Bonus_MICE)/Bonus_MCAR_Data$Bonus_Original*100)
# MAR Error PercentBonus
Bonus_MAR_Error <- tibble(Bonus_SI_Error=abs(Bonus_MAR_Data$Bonus_Original-Bonus_MAR_Data$Bonus_SI)/Bonus_MAR_Data$Bonus_Original*100,
Bonus_kNN_Error=abs(Bonus_MAR_Data$Bonus_Original-Bonus_MAR_Data$Bonus_kNN)/Bonus_MAR_Data$Bonus_Original*100,
Bonus_RF_Error=abs(Bonus_MAR_Data$Bonus_Original-Bonus_MAR_Data$Bonus_RF)/Bonus_MAR_Data$Bonus_Original*100,
Bonus_MICE_Error=abs(Bonus_MAR_Data$Bonus_Original-Bonus_MAR_Data$Bonus_MICE)/Bonus_MAR_Data$Bonus_Original*100)
# MNAR Error PercentBonus
Bonus_MNAR_Error <- tibble(Bonus_SI_Error=abs(Bonus_MNAR_Data$Bonus_Original-Bonus_MNAR_Data$Bonus_SI)/Bonus_MNAR_Data$Bonus_Original*100,
Bonus_kNN_Error=abs(Bonus_MNAR_Data$Bonus_Original-Bonus_MNAR_Data$Bonus_kNN)/Bonus_MNAR_Data$Bonus_Original*100,
Bonus_RF_Error=abs(Bonus_MNAR_Data$Bonus_Original-Bonus_MNAR_Data$Bonus_RF)/Bonus_MNAR_Data$Bonus_Original*100,
Bonus_MICE_Error=abs(Bonus_MNAR_Data$Bonus_Original-Bonus_MNAR_Data$Bonus_MICE)/Bonus_MNAR_Data$Bonus_Original*100)
Bonus_Error <- data.frame(
SI_Error = c(mean(Bonus_MCAR_Error$Bonus_SI_Error) , mean(Bonus_MAR_Error$Bonus_SI_Error) , mean(Bonus_MNAR_Error$Bonus_SI_Error)),
kNN_Error = c(mean(Bonus_MCAR_Error$Bonus_kNN_Error) , mean(Bonus_MAR_Error$Bonus_kNN_Error) , mean(Bonus_MNAR_Error$Bonus_kNN_Error)),
RF_Error = c(mean(Bonus_MCAR_Error$Bonus_RF_Error) , mean(Bonus_MAR_Error$Bonus_RF_Error) , mean(Bonus_MNAR_Error$Bonus_RF_Error)),
MICE_Error = c(mean(Bonus_MCAR_Error$Bonus_MICE_Error), mean(Bonus_MAR_Error$Bonus_MICE_Error), mean(Bonus_MNAR_Error$Bonus_MICE_Error)))
rownames(Bonus_Error) <- c("Bonus_MCAR","Bonus_MAR","Bonus_MNAR")
# Tabulate Errors
Errors_by_Variable <- rbind(JobTitle_Error, Gender_Error, Age_Error, PerfEval_Error, Education_Error, Dept_Error, Seniority_Error, BasePay_Error, Bonus_Error)
Errors_by_Variable
Errors_by_Type <- rbind(JobTitle_Error[1,], Gender_Error[1,], Age_Error[1,], PerfEval_Error[1,], Education_Error[1,], Dept_Error[1,], Seniority_Error[1,], BasePay_Error[1,], Bonus_Error[1,],
JobTitle_Error[2,], Gender_Error[2,], Age_Error[2,], PerfEval_Error[2,], Education_Error[2,], Dept_Error[2,], Seniority_Error[2,], BasePay_Error[2,], Bonus_Error[2,],
JobTitle_Error[3,], Gender_Error[3,], Age_Error[3,], PerfEval_Error[3,], Education_Error[3,], Dept_Error[3,], Seniority_Error[3,], BasePay_Error[3,], Bonus_Error[3,])
Errors_by_Type
# Error Visualization
# Melt data frame into long format
Errors_by_Type_Long = rownames_to_column(Errors_by_Type, "ErrorName")
Errors_by_Type_Long <- melt(Errors_by_Type_Long, id.vars = 'ErrorName', variable.name = 'ImputationType',value.name = "ErrorPercent")
Errors_by_Variable_Long = rownames_to_column(Errors_by_Variable, "ErrorName")
Errors_by_Variable_Long <- melt(Errors_by_Variable_Long, id.vars = 'ErrorName', variable.name = 'ImputationType',value.name = "ErrorPercent")
#create line plot for each column in data frame
df = Errors_by_Variable_Long[c(1:9,28:36,55:63,82:90),]
p1 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Overview of Errors")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("Age|BasePay|Bonus"),ErrorName))
p2 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Continuous Variables")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("Gender"),ErrorName))
p3 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Binary Categorical Variables")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("JobTitle|Dept"),ErrorName))
p4 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Multinomial Categorical Variables")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("PerEval|Education|Seniority"),ErrorName))
p5 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Ordinal Categorical Variables")
df = dplyr::filter(Errors_by_Variable_Long, grepl("MCAR",ErrorName))
p6 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison for MCAR Type of Missing Data")
df = dplyr::filter(Errors_by_Variable_Long, grepl("MAR",ErrorName))
p7 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison for MAR Type of Missing Data")
df = dplyr::filter(Errors_by_Variable_Long, grepl("MNAR",ErrorName))
p8 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison for MNAR Type of Missing Data")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("kNN_Error|RF_Error|MICE_Error"),ImputationType))
p9 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Random Forest and MICE Imputation Methods")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("SI_Error"),ImputationType))
p10 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
scale_fill_manual(values="#F8766D") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Single Imputation Methods")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("kNN_Error"),ImputationType))
p11 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
scale_fill_manual(values="#7CAE00") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between kNN Imputation Methods")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("RF_Error"),ImputationType))
p12 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
scale_fill_manual(values="#00BFC4") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between Random Forest Imputation Methods")
df = dplyr::filter(Errors_by_Variable_Long, grepl(c("MICE_Error"),ImputationType))
p13 = ggplot(df, aes(ErrorName, ErrorPercent, fill=ImputationType)) +
geom_bar(stat = "identity", position = "dodge2") +
scale_fill_manual(values="#C77CFF") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90)) +
labs(title = "Error Comparison between MICE Imputation Methods")
# Visualizations
hist(data)
vis_miss(dataMCAR)
vis_miss(dataMAR)
vis_miss(dataMNAR)
hist(dataMAR_SI)
hist(dataMAR_MICE_Final)
datatable(Errors_by_Type, caption = "Errors by type", filter = "top")
p1
ggarrange(p2,p3,p4,p5,nrow = 2, ncol = 2)
ggarrange(p6,p7,p8,nrow = 2, ncol = 2)
ggarrange(p10,p11,p12,p13,nrow = 2, ncol = 2)