Scribd
Upload
19 views

Exploratory Data Analysis

This document discusses exploratory data analysis of a cancer dataset using various R functions to analyze correlations. Several datasets are extracted from the original cancer data and analyzed using functions for pairwise correlation, scatter plots, normality testing, and simulating raw data from given correlation coefficients. Correlation analysis is performed between variables in each dataset using Pearson and Spearman correlation coefficients. Normality of each variable is also assessed using QQ plots and Wilk-Shapiro tests.

Uploaded by

Cyd Duque
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
19 views

Exploratory Data Analysis

This document discusses exploratory data analysis of a cancer dataset using various R functions to analyze correlations. Several datasets are extracted from the original cancer data and analyzed using functions for pairwise correlation, scatter plots, normality testing, and simulating raw data from given correlation coefficients. Correlation analysis is performed between variables in each dataset using Pearson and Spearman correlation coefficients. Normality of each variable is also assessed using QQ plots and Wilk-Shapiro tests.

Uploaded by

Cyd Duque
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 5

# Mindanao State University

# General Santos City

# Exploratory Data Analysis


# Prepared by: Prof. Carlito O. Daarol
# Math Department
# March 16, 2023

# -------------------------------------------------------------------
# Activate the file containing all functions
# You should modify the file location because it refers to my laptop
# -------------------------------------------------------------------

(drive <- "D:/")


folder_functions <- "Research/thesisfunctions"
filename <- "fn_More_Correlations.R"
source(paste0(drive, folder_functions,"/",filename))

# --------------------------------------------------------
# Using function 1: Read the dataset using function call
# --------------------------------------------------------

# Set pointer to location of my data ( do not use setwd command for data
retrieval)

folder_data <- "C:/Users/Admin/Desktop/Class Lectures/BLecture 0 Graphics in R"


filename <- "Cancer.csv"
data <- readcsv(folder_data,filename)

# Check contents
dim(data)
colnames(data)
head(data)

# ------------------------------------------
# Using R package to display large dataset
# Output visible only if output format is html
# --------------------------- --------------
library(DT)
datatable(data)

# --------------------------------------------------------
# Using function 3: Compute correlation using two columns
# from the dataset
# --------------------------------------------------------

X <- data$breastcancer
Y <- data$co2emissions
corXY <- correlation(X,Y)
corXY
# Result: NA
# This means computation of correlation is not possible because
# of the presence of Missing values

# Possible solution is to omit the NA values


# this is not good because at the end
# X and Y may not have the same length

# Using function 4: put X and Y into 1 dataframe

dataXY <- as.data.frame(cbind(X,Y))


dim(dataXY)

# select only rows with no missing value


dataXY <- na.omit(dataXY)
dim(dataXY)

corXY <- Correcorre(dataXY)


corXY

# Using function 5 and 2: Construct two sets of variables from the data

str(data) # we need to lookup first on the type of variables we have

# select three columns from the data


Set1 <- data[,c("breastcancer", "alcconsumption","internetuserate")]
anyNA(Set1)

# select another three columns from the data


Set2 <- data[,c("co2emissions", "employrate","lifeexpectancy")]
anyNA(Set2)

# If we delete NA values separately then Set1 and Set2 we


# could possibly have unequal rows
# solution is to combine them as dataframe

tmpdat <- cbind(Set1,Set2)


tmpdat <- na.omit(tmpdat)
dim(tmpdat)

# pull out again Set1 and Set2


Set1 <- tmpdat[,1:3]
Set2 <- tmpdat[,4:6]

#process pairwise correlations by feeding the two sets to the 5th function
Pearsonr <- pairwiseCor(Set1,Set2,"pearson")
Spearmanr <- pairwiseCor(Set1,Set2,"spearman")

Pearsonr
Spearmanr

# Table is not good enough for distribution


# Call the function #2 NiceTable to enhance appearance
NiceTable(Pearsonr,"Pearson Correlation Analysis")
NiceTable(Spearmanr,"Spearman Correlation Analysis")

# Using function 6: Compute correlation using only 1 set of data

Pearson1 <- singlesetCor(tmpdat,"pearson")


Spearman1 <- singlesetCor(tmpdat,"spearman")

#display unformatted table


Pearson1
Spearman1

# Display a better table


NiceTable(Pearson1, "Pearson Correlation Analysis")
NiceTable(Spearman1, "Pearson Correlation Analysis")

# Using function 7: Correlation Coefficents in table format

CorrsjPlot(Set1,"pearson","Pearson Correlation Coefficients")


CorrsjPlot(Set2,"pearson","Pearson Correlation Coefficients")

# Using function 8: Scatter Plot

Set1name <- colnames(Set1)


CorrePlotXY(Set1,Set1name[1],Set1name[2],"blue", "XAxis", "YAxis","pearson")

Set2name <- colnames(Set2)


CorrePlotXY(Set2,Set2name[1],Set2name[2],"blue", "XAxis", "YAxis","pearson")

# Use double for loop to generate all plots for Set 1


for (i in 1:(ncol(Set1)-1)) {
for (j in (i+1):ncol(Set1)){
print(CorrePlotXY(Set1,Set1name[i],Set1name[j],"blue", Set1name[i],
Set1name[j],"pearson"))

}
}

# Use double for loop to generate all plots for Set 2


for (i in 1:(ncol(Set2)-1)) {
for (j in (i+1):ncol(Set2)){
print(CorrePlotXY(Set2,Set2name[i],Set2name[j],"blue", Set2name[i],
Set2name[j],"pearson"))
}
}

#Remark: Plots under double for loop will not appear without the pront command

# Using function 9: How to verify if the data


# satisfies the normal distribution using Wilk-Shapiro test
# Using function 10: How to verify if the data
# satisfies the normal distribution using graphs

NiceTable(Set1,"Dataset in wide original format")

# convert data to long format first


data_long <- melt(Set1)
NiceTable(data_long,"Dataset in long format")
QQNormality_Plot(data_long)

# Points must fall inside the confidence band


# for it to be called as normally distributed.
# If not satisfied call the distribution as Non-normal (skewed)

# Using function 10: How to verify if the data


# satisfies the normal distribution using graphs

NiceTable(Set1,"Dataset in wide original format")

# convert data to long format first


data_long <- melt(Set1)
NiceTable(data_long,"Dataset in long format")
QQNormality_Plot(data_long)

# Points must fall inside the confidence band


# for it to be called as normally distributed.
# If not satisfied call the distribution as Non-normal (skewed)

# Using function 11: For a given set of correlation coefficients, Generate the
# corresponding raw data X and Y.

PlotHistDensity(Set1)

# Using function 12: For a given set of correlation coefficients, Generate the
# corresponding raw data X and Y.

sampleCor <- c(0.214, 0.4, 0.617, 0.742, 0.851, 0.915)


Simulate_XY_From_Correlations(sampleCor)

sampleCor <- c(0.214, 0.3, 0.617, 0.76, 0.851, 0.915)


Simulate_XY_From_Correlations(sampleCor)

# View generated data


gendata <- read.csv("DatXY.csv")
NiceTable(gendata,"Generated Datasets")

You might also like