Factors affecting the Monthly Income of the employee
Mohammad Fuaad Khan
January 30, 2018
This analysis is submitted as the Capstone Project work done for the internship pursued under the guidance of Prof.Sameer Mathur for duration Jan 1,2018-Jan 27,2018.
1. Introduction
There is a phrase quoted by Robert Heinlein in his sci-fi novel: “There is no such thing as a free lunch” Indeed, this has proven to be completely correct in reality too because income has become the prime motive for any person to do any work or provide any service. Every person expects the compensation for his hard work. But, since everybody expects compensation, so there are needs for certain criteria that are efficient to judge the extent to which the job has been successfully completed, both qualitatively as well as quantitatively. And after the performance has been carefully analyzed, it is the time to reward back the employee in order to ensure better job retention and enthusiastic response for the future projects. But, this assurance is very subjective as what is worthy for one can be peanuts for other. Hence, there is a need to justify the extent of effort that employee has put-in. This is taken care generally during the performance appraisal of the employee in which different parameters are taken into consideration for rating the employee. But how to quantify these parameters and judge the performance extent? This is done by collecting the data of all the employees and carrying out analysis on different parameters so as to know who deserves what. Besides, this analysis can also be helpful is estimating the underpaid or overpaid employees and appraise them accordingly. Apart from that, it may also help in monetary budgeting as well as in offering the packages to the new hirees as per their worth. The project is centered on how this analysis which is carried over the past data, can be helpful to study the dependency of monthly income on different parameters and estimate it by incorporating statistics.
2. Overview of the Study
Our field of study is concerned with the people that are working in different companies and their job related attributes. The specific objective of this study was to investigate the different factors governing monthly income of an employee. This study analyzes the dependency of monthly income on different parameters and demonstrates the dependency pattern as well. Our goal was to find whether monthly income actually is a function of these parameters or not. The study parameters selected for carrying out the analysis are discussed in the subsequent section of the report.
2.1 Data
The data used for carrying out the analysis was an open source data which is available on the IBM Watson Analytics website. The URL for the website is given as https://www.ibm.com/communities/analytics/watson-analytics-blog/hr-employee-attrition/ We have focused on analyzing the extent to which the monthly income is governed by different parameters. The detailed description of each of the parameters is as follows: i) Gender: This is a qualitative variable which differentiates the employee based on the gender to which they belong. In the dataset, field Gender accounts for this information which has Female and Male as entries that correspond to their usual meanings.
Department: This is a qualitative variable which differentiates the employee based on the department to which they belong. In the dataset, field Department accounts for this information which has Human Resources, Research & Development and Sales as entries that correspond to their usual meanings.
Business Travel Status: This is a qualitative variable which differentiates the employee based on the extent to which they travel. In the dataset, field BusinessTravel accounts for this information which has Travel_Rarely, Travel_Frequently and Non-Travel as entries that correspond to employees who travel rarely, travel frequently and do not travel respectively.
Marital Status: This is a qualitative variable which differentiates the employee based on their marital status. In the dataset, field MaritalStatus accounts for this information which has Single, Married and Divorced as entries that correspond to their usual meanings.
Job Role: This is a qualitative variable which differentiates the employee based on their roles in the company. In the dataset, field JobRole accounts for this information which has Healthcare Representatives, Human Resources, Laboratory Technician, Manager, Manufacturing Director, Research Director, Research Scientist, Sales Executive and Sales Representative as entries that correspond to their usual meanings.
Age: This is a quantitative variable which records the data corresponding to the age of the employee. In the dataset, field Age accounts for this information which has the respective age of the employees as its entries.
Education Level: This is a quantitative variable which records the data corresponding to the education level of the employee. In the dataset, field Age accounts for this information which have been collected over the Likert Scale over the range 1-5.
1 Below College 2 College 3 Bachelor 4 Master 5 Doctor
Job Level: This is a quantitative variable which records the data corresponding to the job level of the employee. In the dataset, field JobLevel accounts for this information which have been collected over the Likert Scale over the range 1-5.
1 Staff 2 Representative 3 Manager 4 Executive 5 Director
Number of companies previously worked: This is a quantitative variable which records the data corresponding to the number of companies previously served by the employee. In the dataset, field NumCompaniesWorked accounts for this information which has the respective value of number of companies an employee has worked for, as its entries.
Percentage of salary hike obtained: This is a quantitative variable which records the data corresponding to the percentage salary hike received by the employee. In the dataset, field PercentSalaryHike accounts for this information which has the respective value of percentage salary hike an employee has received, as its entries.
Total Work Experience: This is a quantitative variable which records the data corresponding to the total work experience possessed by the employee. In the dataset, field TotalWorkingYears accounts for this information which has the respective value of no. of years an employee has experience of working, as its entries.
Years working in the company: This is a quantitative variable which records the data corresponding to the total number of years the employee has served the company. In the dataset, field YearsAtCompany accounts for this information which has the respective value of no. of years an employee has served the company, as its entries.
2.2 Hypothesis
The main intention of the analysis is to find whether Monthly Income of an employee is significantly dependent upon the aforementioned parameters or not. As every analysis commence by considering a null hypothesis, the hypothesis in this case is defined as: H0: The average monthly income of the employee is independent of all the parameters. H1: The average monthly income of the employee is dependent on the parameters.
2.3 Model
The regression model was created for the following analysis and it was found that of all the 13 parameters, only 5 are significant. The regression equation for the model thus developed can be given as:
Monthly Income = (-111.39) + 90 Gender + (451.6) BusinessTravelTravel_Frequently+ (276.33) BusinessTravelTravel_Rarely - (237.67) JobRoleHuman Resources - (900.24) JobRoleLaboratory Technician + (4022.03) JobRoleManager - (356.7) JobRoleManufacturing Director + (3791.69) JobRoleResearch Director - (509.37) JobRoleResearch Scientist - (252.52) JobSales Executive - (775.40) JobRoleSales Representative + (2779.29) JobLevel + (36.43) TotalWorkingYears
# Reading the file Monthly Income Project_data.csv
monthinc.df <- read.csv(paste("Monthly Income Project_data.csv", sep= ""))
View(monthinc.df)
# Regression Analysis
regress3 <- lm(MonthlyIncome ~ Gender + BusinessTravel + JobRole + JobLevel + TotalWorkingYears , data = monthinc.df)
summary(regress3)##
## Call:
## lm(formula = MonthlyIncome ~ Gender + BusinessTravel + JobRole +
## JobLevel + TotalWorkingYears, data = monthinc.df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3644.3 -645.1 53.5 660.8 4079.7
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -111.387 268.895 -0.414 0.67883
## GenderMale 89.988 83.764 1.074 0.28307
## BusinessTravelTravel_Frequently 451.610 167.411 2.698 0.00716 **
## BusinessTravelTravel_Rarely 276.331 145.512 1.899 0.05798 .
## JobRoleHuman Resources -237.667 295.366 -0.805 0.42130
## JobRoleLaboratory Technician -900.236 195.818 -4.597 5.10e-06 ***
## JobRoleManager 4022.027 248.321 16.197 < 2e-16 ***
## JobRoleManufacturing Director -356.711 193.798 -1.841 0.06611 .
## JobRoleResearch Director 3791.686 250.255 15.151 < 2e-16 ***
## JobRoleResearch Scientist -509.365 195.868 -2.601 0.00951 **
## JobRoleSales Executive -252.517 168.092 -1.502 0.13349
## JobRoleSales Representative -775.396 241.189 -3.215 0.00137 **
## JobLevel 2779.291 96.307 28.859 < 2e-16 ***
## TotalWorkingYears 36.430 8.802 4.139 3.93e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1080 on 686 degrees of freedom
## Multiple R-squared: 0.9472, Adjusted R-squared: 0.9462
## F-statistic: 946.6 on 13 and 686 DF, p-value: < 2.2e-162.4 Results
The analysis was carried out in order to study the dependency how Monthly Income of the employee is dependent on different variables. The following insights were drawn out of it: 1) Monthly Income of the employee depends very significantly on Job Role and Job Level of the employee. 2) Monthly Income also has significant dependency on Gender, Business Travel Status and Total Work Experience of the employee. 3) Monthly Income has weak dependency on Department and Marital Status of the employee. 4) Monthly Income is almost independent of Age and Education Level of the employee, Number of companies that the employee has served earlier, Percent Salary Hike obtained by employee and Years the employee has served the company. 5) Job Level contributes highly to Monthly Income implying that a raise of about 2780 is obtained for each increment in the Job Level. 6) Monthly income is found to be maximum for Manager and Research Director and minimum for Laboratory Technician and Sales Representative among all the other job roles. 7) Monthly income is having a difference of just about 90 if the employee is male rather than female ensuring gender equality in the company. 8) Monthly income of employee who travel frequently exceeds by about 175 than one who travel rarely. 9) Monthly income of employee rises by about 36 per unit rise in the Total work experience of the employee.
3. Conclusion
This analysis was motivated by the need to explore the dependency of monthly income on different parameters so that we could improve our understanding of how it is influenced through different study parameters and how it can be correctly estimated for a variety of employees having different attributes and records. The unique contribution of this paper is that the world believes in the gender equality as no significant dependency was found between monthly income and gender of the employee. We also found that total work experience also does not significantly contribute to increase in the monthly income. And its quiet obvious that as the Job Level will increase, the monthly income will rise which can also be concluded from the analysis as well.
4. References
- https://www.ibm.com/communities/analytics/watson-analytics-blog/hr-employee-attrition/
- https://en.wikipedia.org/wiki/There_ain%27t_no_such_thing_as_a_free_lunch
- https://www.udemy.com/data-science-and-analytics-using-r/learn/v4/overview
- https://drive.google.com/file/d/1PemBkTzSW7F8CJhWTv0j2B23rF_PCU1r/view
- https://drive.google.com/file/d/19OqShylgDvHGrtIakmPJdY7LCOXiXH_z/view
- https://docs.google.com/document/d/1Rh1GTQZObHkuqoku_2FlTTRjSRX-pYqziK7xPa03yEI/edit
- https://docs.google.com/document/d/1lM3io06gUZMK97NoOcFWSVnUdQqd9hp_Ef2FVbSREE0/edit
- https://docs.google.com/document/d/1zaKJFXUaEDcprhcw0crMDd9KJolCHq30ehDGpsgMJd4/edit
- https://drive.google.com/file/d/0B3oXSJHLqZqXMGZGTzlUSjE2UTQ/view
- https://drive.google.com/file/d/0B3oXSJHLqZqXUWI1WDVXamhoU3c/view
5. Appendix
Reading the file Monthly Income Project_data.csv and measuring dimensions
monthinc.df <- read.csv(paste("Monthly Income Project_data.csv", sep= ""))
View(monthinc.df)
dim(monthinc.df)## [1] 700 14Generating summary for the dataframe
library(psych)
describe(monthinc.df)## vars n mean sd median trimmed mad min
## Employee.ID 1 700 1477.59 280.39 1469.5 1475.78 355.08 1001
## Gender* 2 700 1.59 0.49 2.0 1.61 0.00 1
## Department* 3 700 2.27 0.52 2.0 2.26 0.00 1
## BusinessTravel* 4 700 2.63 0.64 3.0 2.78 0.00 1
## MaritalStatus* 5 700 2.09 0.74 2.0 2.12 1.48 1
## JobRole* 6 700 5.54 2.43 6.0 5.70 2.97 1
## Age 7 700 36.83 9.32 35.5 36.36 9.64 18
## Education 8 700 2.88 1.03 3.0 2.94 1.48 1
## JobLevel 9 700 2.05 1.10 2.0 1.88 1.48 1
## NumCompaniesWorked 10 700 2.71 2.52 2.0 2.38 1.48 0
## PercentSalaryHike 11 700 15.28 3.72 14.0 14.88 2.97 11
## TotalWorkingYears 12 700 11.20 7.80 9.5 10.26 5.19 0
## YearsAtCompany 13 700 6.98 6.37 5.0 5.84 4.45 0
## MonthlyIncome 14 700 6419.27 4654.16 4887.0 5581.97 3220.95 1009
## max range skew kurtosis se
## Employee.ID 1976 975 0.05 -1.16 10.60
## Gender* 2 1 -0.37 -1.86 0.02
## Department* 3 2 0.24 -0.50 0.02
## BusinessTravel* 3 2 -1.53 1.01 0.02
## MaritalStatus* 3 2 -0.15 -1.16 0.03
## JobRole* 9 8 -0.39 -1.14 0.09
## Age 60 42 0.44 -0.43 0.35
## Education 5 4 -0.25 -0.64 0.04
## JobLevel 5 4 1.05 0.50 0.04
## NumCompaniesWorked 9 9 1.03 0.01 0.10
## PercentSalaryHike 25 14 0.77 -0.48 0.14
## TotalWorkingYears 40 40 1.18 1.18 0.29
## YearsAtCompany 40 40 1.90 4.43 0.24
## MonthlyIncome 19999 18990 1.39 1.08 175.91Creating one-way contingency tables for Gender and displaying the distribution on bar graph
table(monthinc.df$Gender)##
## Female Male
## 286 414counts_gen <-table(monthinc.df$Gender)
barplot(counts_gen, width=1, space=1, main = "Gender Distribution", xlab="Gender",col=c( "pink","navy"),
ylim=c(0,450), xlim=c(0,10), names.arg=c("Female","Male"))
Axis(side=1, labels=FALSE)
Creating one-way contingency tables for Department and displaying the distribution on bar graph
table(monthinc.df$Department)##
## Human Resources Research & Development Sales
## 24 460 216counts_dep <-table(monthinc.df$Department)
barplot(counts_dep, width=1, space=1, main = "Department Distribution", xlab="Department",
col=c("red","gold","purple"), ylim=c(0,500), xlim=c(0,10), names.arg=c("HR","R&D","Sales"))
Axis(side=1, labels=FALSE)
Creating one-way contingency tables for Business Travelling and displaying the distribution on bar graph
table(monthinc.df$BusinessTravel)##
## Non-Travel Travel_Frequently Travel_Rarely
## 64 129 507counts_bustr <-table(monthinc.df$BusinessTravel)
barplot(counts_bustr, width=1, space=1, main = "Business Travelling Status", xlab="Status",
col=c("wheat","green1","tomato"), ylim=c(0,500), xlim=c(0,10),
names.arg=c(" Don't Travel","Frequently","Rarely"))
Axis(side=1, labels=FALSE)
Creating one-way contingency tables for Marital Status and displaying the distribution on bar graph
table(monthinc.df$MaritalStatus)##
## Divorced Married Single
## 161 312 227counts_marstat <-table(monthinc.df$MaritalStatus)
barplot(counts_marstat, width=1, space=1, main = "Marital Status", xlab="Status",
col=c("white","firebrick","skyblue"), ylim=c(0,500), xlim=c(0,10),
names.arg=c("Divorced","Married","Single"))
Axis(side=1, labels=FALSE)
Creating one-way contingency tables for Job Role and displaying the distribution on bar graph
table(monthinc.df$JobRole)##
## Healthcare Representative Human Resources
## 58 19
## Laboratory Technician Manager
## 123 51
## Manufacturing Director Research Director
## 69 39
## Research Scientist Sales Executive
## 142 155
## Sales Representative
## 44counts_jobr <-table(monthinc.df$JobRole)
barplot(counts_jobr, width=1, space=1, main = "Job Role", xlab="Role", ylim=c(0,380), xlim=c(0,20), xaxt="n", col=c("violet","ivory","blue","green","yellow","darkorange","red","cyan","magenta"))
legend("topleft", c("Health Rep.","HR","Lab. Tech.","Manager","Manufac. Director","Research Director", "Research Scientist","Sales Executive","Sales Representative"), fill=c("violet","ivory","blue","green", "yellow","darkorange","red","cyan","magenta"))
Axis(side=1, labels=FALSE)
Creating two-way contingency tables for Gender & Department
two_way_tab <-xtabs(~ Gender + Department, data = monthinc.df)
addmargins(two_way_tab)## Department
## Gender Human Resources Research & Development Sales Sum
## Female 8 185 93 286
## Male 16 275 123 414
## Sum 24 460 216 700Creating two-way contingency tables for Job Role & Business Travel Status
two_way_tab <-xtabs(~ JobRole + BusinessTravel, data = monthinc.df)
addmargins(two_way_tab)## BusinessTravel
## JobRole Non-Travel Travel_Frequently Travel_Rarely Sum
## Healthcare Representative 7 12 39 58
## Human Resources 0 2 17 19
## Laboratory Technician 13 26 84 123
## Manager 6 7 38 51
## Manufacturing Director 7 12 50 69
## Research Director 2 5 32 39
## Research Scientist 7 23 112 142
## Sales Executive 21 30 104 155
## Sales Representative 1 12 31 44
## Sum 64 129 507 700Creating two-way contingency tables for Gender & Marital Status
two_way_tab <-xtabs(~ Gender + MaritalStatus, data = monthinc.df)
addmargins(two_way_tab)## MaritalStatus
## Gender Divorced Married Single Sum
## Female 61 122 103 286
## Male 100 190 124 414
## Sum 161 312 227 700Creating two-way contingency tables for Department & Business Travel Status
two_way_tab <-xtabs(~ Department + BusinessTravel, data = monthinc.df)
addmargins(two_way_tab)## BusinessTravel
## Department Non-Travel Travel_Frequently Travel_Rarely Sum
## Human Resources 2 3 19 24
## Research & Development 39 83 338 460
## Sales 23 43 150 216
## Sum 64 129 507 700Creating three-way contingency tables for Department & Business Travel Status
three_way_tab <- xtabs(~ Department + JobRole + Gender, data = monthinc.df)
ftable(three_way_tab)## Gender Female Male
## Department JobRole
## Human Resources Healthcare Representative 0 0
## Human Resources 6 13
## Laboratory Technician 0 0
## Manager 2 3
## Manufacturing Director 0 0
## Research Director 0 0
## Research Scientist 0 0
## Sales Executive 0 0
## Sales Representative 0 0
## Research & Development Healthcare Representative 27 31
## Human Resources 0 0
## Laboratory Technician 41 82
## Manager 12 17
## Manufacturing Director 32 37
## Research Director 15 24
## Research Scientist 58 84
## Sales Executive 0 0
## Sales Representative 0 0
## Sales Healthcare Representative 0 0
## Human Resources 0 0
## Laboratory Technician 0 0
## Manager 11 6
## Manufacturing Director 0 0
## Research Director 0 0
## Research Scientist 0 0
## Sales Executive 63 92
## Sales Representative 19 25Box plot for Age
boxplot(monthinc.df$Age, horizontal = TRUE, main = "Box Plot for Age", xlab = "Age", col = "chocolate")
Box plot for Work Experience
boxplot(monthinc.df$TotalWorkingYears, horizontal = TRUE, main = "Box Plot for Work Experience", xlab = "Years", col = "gold")
Box plot for Monthly Income
boxplot(monthinc.df$MonthlyIncome, horizontal = TRUE, main = "Box Plot for Monthly Income", xlab = "Amount", col = "red")
Histogram for Percentage hike in the Salary
table(monthinc.df$PercentSalaryHike)##
## 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
## 103 95 94 92 47 39 34 45 37 25 25 29 17 12 6hist(monthinc.df$PercentSalaryHike,
main="Variation in Percentage hike in the Salary",
xlab="Percentage",
ylab="Count",
xlim=c(10,30), ylim=c(0,270),
breaks=5,
col=c("orange", "darkorange", "orangered", "orangered3", "red", "red3", "firebrick3", "firebrick")) 
Histogram for Years in the company
table(monthinc.df$YearsAtCompany)##
## 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 25
## 21 87 61 63 50 96 36 42 39 39 53 15 5 12 6 7 5 6 7 2 12 6 7 5 3
## 26 27 29 31 32 33 34 36 37 40
## 2 2 1 3 1 2 1 1 1 1hist(monthinc.df$YearsAtCompany,
main="No. of years in the company",
xlab="Years",
ylab="Count",
xlim=c(0,40), ylim=c(0,300),
breaks=10,
col=c("lightblue", "lightblue3", "royalblue", "blue", "royalblue4", "blue3", "dark blue", "navy")) 
Correlation matrix and Corrgram Generation
round(cor(monthinc.df[,7:14]), 3)## Age Education JobLevel NumCompaniesWorked
## Age 1.000 0.196 0.495 0.299
## Education 0.196 1.000 0.136 0.120
## JobLevel 0.495 0.136 1.000 0.121
## NumCompaniesWorked 0.299 0.120 0.121 1.000
## PercentSalaryHike -0.054 -0.030 -0.096 -0.004
## TotalWorkingYears 0.675 0.157 0.780 0.220
## YearsAtCompany 0.304 0.086 0.566 -0.137
## MonthlyIncome 0.484 0.134 0.951 0.123
## PercentSalaryHike TotalWorkingYears YearsAtCompany
## Age -0.054 0.675 0.304
## Education -0.030 0.157 0.086
## JobLevel -0.096 0.780 0.566
## NumCompaniesWorked -0.004 0.220 -0.137
## PercentSalaryHike 1.000 -0.096 -0.085
## TotalWorkingYears -0.096 1.000 0.640
## YearsAtCompany -0.085 0.640 1.000
## MonthlyIncome -0.093 0.763 0.543
## MonthlyIncome
## Age 0.484
## Education 0.134
## JobLevel 0.951
## NumCompaniesWorked 0.123
## PercentSalaryHike -0.093
## TotalWorkingYears 0.763
## YearsAtCompany 0.543
## MonthlyIncome 1.000library(corrgram)## Warning: replacing previous import by 'magrittr::%>%' when loading
## 'dendextend'corrgram(monthinc.df, order=TRUE, lower.panel=panel.shade,upper.panel=panel.pie, text.panel=panel.txt,main="Corrgram for different variables")
ScatterPlot Martrix for different variables
library(car)##
## Attaching package: 'car'## The following object is masked from 'package:psych':
##
## logitscatterplotMatrix(~ Gender+ Department + MonthlyIncome + MaritalStatus + Age, data=monthinc.df,
main="Variation of Monthly Income with Gender, Department,
Marital Status and Age")
scatterplotMatrix(~ Education + JobLevel + MonthlyIncome + TotalWorkingYears + YearsAtCompany , data=monthinc.df, main="Variation of Monthly Income with Education, Job Level,
Total Work Experience and Years in the Company")
Defining Null Hypothesis
Considering the null hypothesis H0: There is no significant difference in the value of both means. H1: There is a significant difference in the value of both means.
t-Test for dependency of Monthly Income on Age
t.test(monthinc.df$MonthlyIncome, monthinc.df$Age)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$Age
## t = 36.282, df = 699.01, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6037.062 6727.818
## sample estimates:
## mean of x mean of y
## 6419.27429 36.83429-> Since p-value < 2.2e-16, so Age is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis
t-Test for dependency of Monthly Income on Education
t.test(monthinc.df$MonthlyIncome, monthinc.df$Education)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$Education
## t = 36.475, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6071.017 6761.771
## sample estimates:
## mean of x mean of y
## 6419.274 2.880-> Since p-value < 2.2e-16, so Education is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis
t-Test for dependency of Monthly Income on Job Level
t.test(monthinc.df$MonthlyIncome, monthinc.df$JobLevel)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$JobLevel
## t = 36.48, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6071.850 6762.604
## sample estimates:
## mean of x mean of y
## 6419.274286 2.047143-> Since p-value < 2.2e-16, so Job Level is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
t-Test for dependency of Monthly Income on Number of companies worked earlier
t.test(monthinc.df$MonthlyIncome, monthinc.df$NumCompaniesWorked)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$NumCompaniesWorked
## t = 36.476, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6071.187 6761.941
## sample estimates:
## mean of x mean of y
## 6419.274 2.710-> Since p-value < 2.2e-16, so Number of companies worked earlier is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
t-Test for dependency of Monthly Income on Percentage Hike in salary
t.test(monthinc.df$MonthlyIncome, monthinc.df$PercentSalaryHike)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$PercentSalaryHike
## t = 36.405, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6058.619 6749.373
## sample estimates:
## mean of x mean of y
## 6419.27429 15.27857-> Since p-value < 2.2e-16, so Percentage Hike in salary is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
t-Test for dependency of Monthly Income on Total Work Experience
t.test(monthinc.df$MonthlyIncome, monthinc.df$TotalWorkingYears)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$TotalWorkingYears
## t = 36.428, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6062.698 6753.453
## sample estimates:
## mean of x mean of y
## 6419.27429 11.19857-> Since p-value < 2.2e-16, so Total Work Experience is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
t-Test for dependency of Monthly Income on Years in the present company
t.test(monthinc.df$MonthlyIncome, monthinc.df$YearsAtCompany)##
## Welch Two Sample t-test
##
## data: monthinc.df$MonthlyIncome and monthinc.df$YearsAtCompany
## t = 36.452, df = 699, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 6066.918 6757.673
## sample estimates:
## mean of x mean of y
## 6419.274286 6.978571-> Since p-value < 2.2e-16, so Total Work Experience is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
Converting variables into factors
# Converting Gender into factor variable
monthinc.df$Gender[monthinc.df$Res == 0] <- 'Female'
monthinc.df$Gender[monthinc.df$Res == 1] <- 'Male'
monthinc.df$Gender<- factor(monthinc.df$Gender)
# Converting Department into factor variable
monthinc.df$Department[monthinc.df$Res == 1] <- 'Human Resources'
monthinc.df$Department[monthinc.df$Res == 2] <- 'Research & Development'
monthinc.df$Department[monthinc.df$Res == 3] <- 'Sales'
monthinc.df$Department<- factor(monthinc.df$Department)
# Converting Business Travel Status into factor variable
monthinc.df$BusinessTravel[monthinc.df$Res == 0] <- 'Non-Travel'
monthinc.df$BusinessTravel[monthinc.df$Res == 1] <- 'Travel_Frequently'
monthinc.df$BusinessTravel[monthinc.df$Res == 2] <- 'Travel_Rarely'
monthinc.df$BusinessTravel<- factor(monthinc.df$BusinessTravel)
# Converting Marital Status into factor variable
monthinc.df$MaritalStatus[monthinc.df$Res == 0] <- 'Divorced'
monthinc.df$MaritalStatus[monthinc.df$Res == 1] <- 'Married'
monthinc.df$MaritalStatus[monthinc.df$Res == 2] <- 'Single'
monthinc.df$MaritalStatus<- factor(monthinc.df$MaritalStatus)
# Converting Job Role into factor variable
monthinc.df$JobRole[monthinc.df$Res == 1] <- 'Healthcare Representative'
monthinc.df$JobRole[monthinc.df$Res == 2] <- 'Human Resources'
monthinc.df$JobRole[monthinc.df$Res == 3] <- 'Laboratory Technician'
monthinc.df$JobRole[monthinc.df$Res == 4] <- 'Manager'
monthinc.df$JobRole[monthinc.df$Res == 5] <- 'Manufacturing Director'
monthinc.df$JobRole[monthinc.df$Res == 6] <- 'Research Director'
monthinc.df$JobRole[monthinc.df$Res == 7] <- 'Research Scientist'
monthinc.df$JobRole[monthinc.df$Res == 8] <- 'Sales Executive'
monthinc.df$JobRole[monthinc.df$Res == 9] <- 'Sales Representative'
monthinc.df$JobRole<- factor(monthinc.df$JobRole)
# Adding MonthlyIncome_Range variable
monthinc.df$MonthlyIncome_Range<-cut(monthinc.df$MonthlyIncome, seq(0,20000,4000), right=FALSE, labels=c(1:5))Chi-square Test for dependency of Monthly Income on Gender
chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df$Gender)##
## Pearson's Chi-squared test
##
## data: monthinc.df$MonthlyIncome_Range and monthinc.df$Gender
## X-squared = 9.9068, df = 4, p-value = 0.04203-> Since p-value = 0.04203, so Gender is a significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
Chi-square Test for dependency of Monthly Income on Department
chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df$Department)## Warning in chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df
## $Department): Chi-squared approximation may be incorrect##
## Pearson's Chi-squared test
##
## data: monthinc.df$MonthlyIncome_Range and monthinc.df$Department
## X-squared = 63.027, df = 8, p-value = 1.183e-10-> Since p-value = 1.183e-10, so Department is a very significant contributor in estimating Monthly Income. Hence we reject the null hypothesis.
Chi-square Test for dependency of Monthly Income on Business Travel Status
chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df$BusinessTravel)## Warning in chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df
## $BusinessTravel): Chi-squared approximation may be incorrect##
## Pearson's Chi-squared test
##
## data: monthinc.df$MonthlyIncome_Range and monthinc.df$BusinessTravel
## X-squared = 14.648, df = 8, p-value = 0.06636-> Since p-value = 0.06636, so Business Travel Status is not a significant contributor in estimating Monthly Income. Hence we accept the null hypothesis.
Chi-square Test for dependency of Monthly Income on Marital Status
chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df$MaritalStatus)##
## Pearson's Chi-squared test
##
## data: monthinc.df$MonthlyIncome_Range and monthinc.df$MaritalStatus
## X-squared = 8.6358, df = 8, p-value = 0.3739-> Since p-value = 0.3739, so Marital Status is not a significant contributor in estimating Monthly Income. Hence we accept the null hypothesis.
Chi-square Test for dependency of Monthly Income on Job Role
chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df$JobRole)## Warning in chisq.test(monthinc.df$MonthlyIncome_Range, monthinc.df
## $JobRole): Chi-squared approximation may be incorrect##
## Pearson's Chi-squared test
##
## data: monthinc.df$MonthlyIncome_Range and monthinc.df$JobRole
## X-squared = 1076, df = 32, p-value < 2.2e-16-> Since p-value < 2.2e-16, so Job Role is a very significant contributor in estimating Monthly Income. Hence we accept the null hypothesis.
Regression Analysis
# Model 1
regress1 <- lm(MonthlyIncome ~ Gender + Department + BusinessTravel + MaritalStatus + JobRole + Age + Education + JobLevel + NumCompaniesWorked + PercentSalaryHike + TotalWorkingYears + YearsAtCompany , data = monthinc.df)
summary(regress1)##
## Call:
## lm(formula = MonthlyIncome ~ Gender + Department + BusinessTravel +
## MaritalStatus + JobRole + Age + Education + JobLevel + NumCompaniesWorked +
## PercentSalaryHike + TotalWorkingYears + YearsAtCompany, data = monthinc.df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3650.8 -640.1 39.5 668.0 4086.6
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -300.8557 668.5264 -0.450 0.65283
## GenderMale 88.5381 84.9566 1.042 0.29771
## DepartmentResearch & Development 38.5753 529.7776 0.073 0.94198
## DepartmentSales -323.8944 559.7732 -0.579 0.56304
## BusinessTravelTravel_Frequently 448.5799 169.7570 2.642 0.00842 **
## BusinessTravelTravel_Rarely 275.4782 147.7588 1.864 0.06270 .
## MaritalStatusMarried 58.2552 106.3658 0.548 0.58409
## MaritalStatusSingle 61.1117 113.6017 0.538 0.59079
## JobRoleHuman Resources -196.3777 615.0775 -0.319 0.74962
## JobRoleLaboratory Technician -903.0242 197.8531 -4.564 5.96e-06 ***
## JobRoleManager 4164.8585 286.3968 14.542 < 2e-16 ***
## JobRoleManufacturing Director -351.4553 195.0757 -1.802 0.07205 .
## JobRoleResearch Director 3807.5247 252.7972 15.062 < 2e-16 ***
## JobRoleResearch Scientist -511.6669 198.0583 -2.583 0.00999 **
## JobRoleSales Executive 112.4187 380.5108 0.295 0.76775
## JobRoleSales Representative -412.8890 414.7542 -0.996 0.31985
## Age 1.0651 6.2762 0.170 0.86529
## Education 4.3941 41.2531 0.107 0.91520
## JobLevel 2771.3820 98.2868 28.197 < 2e-16 ***
## NumCompaniesWorked 4.9890 18.3792 0.271 0.78613
## PercentSalaryHike 4.0238 11.2243 0.358 0.72009
## TotalWorkingYears 35.9719 11.6733 3.082 0.00214 **
## YearsAtCompany -0.1172 9.3219 -0.013 0.98997
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1085 on 677 degrees of freedom
## Multiple R-squared: 0.9473, Adjusted R-squared: 0.9456
## F-statistic: 553.5 on 22 and 677 DF, p-value: < 2.2e-16# Model 2
regress2 <- lm(MonthlyIncome ~ Gender + BusinessTravel + JobRole + Age + Education + JobLevel + NumCompaniesWorked + PercentSalaryHike + TotalWorkingYears , data = monthinc.df)
summary(regress2)##
## Call:
## lm(formula = MonthlyIncome ~ Gender + BusinessTravel + JobRole +
## Age + Education + JobLevel + NumCompaniesWorked + PercentSalaryHike +
## TotalWorkingYears, data = monthinc.df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3637.0 -646.7 53.5 660.9 4100.8
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -217.4699 379.9958 -0.572 0.567310
## GenderMale 91.8727 84.4311 1.088 0.276918
## BusinessTravelTravel_Frequently 450.4878 168.7186 2.670 0.007765 **
## BusinessTravelTravel_Rarely 276.2277 146.3652 1.887 0.059551 .
## JobRoleHuman Resources -240.6726 296.6887 -0.811 0.417536
## JobRoleLaboratory Technician -897.0046 196.6876 -4.561 6.05e-06 ***
## JobRoleManager 4028.3348 249.4139 16.151 < 2e-16 ***
## JobRoleManufacturing Director -356.6867 194.4279 -1.835 0.067009 .
## JobRoleResearch Director 3791.7969 251.3978 15.083 < 2e-16 ***
## JobRoleResearch Scientist -504.6747 196.6720 -2.566 0.010498 *
## JobRoleSales Executive -249.5603 168.6999 -1.479 0.139518
## JobRoleSales Representative -767.9924 242.7296 -3.164 0.001625 **
## Age 1.3406 6.1411 0.218 0.827260
## Education 0.6416 40.8904 0.016 0.987486
## JobLevel 2782.7533 96.8472 28.733 < 2e-16 ***
## NumCompaniesWorked 5.0694 17.2895 0.293 0.769451
## PercentSalaryHike 3.1814 11.1274 0.286 0.775034
## TotalWorkingYears 34.7833 10.2763 3.385 0.000753 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1083 on 682 degrees of freedom
## Multiple R-squared: 0.9472, Adjusted R-squared: 0.9459
## F-statistic: 719.9 on 17 and 682 DF, p-value: < 2.2e-16# Model 3
regress3 <- lm(MonthlyIncome ~ Gender + BusinessTravel + JobRole + JobLevel + TotalWorkingYears ,
data = monthinc.df)
summary(regress3)##
## Call:
## lm(formula = MonthlyIncome ~ Gender + BusinessTravel + JobRole +
## JobLevel + TotalWorkingYears, data = monthinc.df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3644.3 -645.1 53.5 660.8 4079.7
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -111.387 268.895 -0.414 0.67883
## GenderMale 89.988 83.764 1.074 0.28307
## BusinessTravelTravel_Frequently 451.610 167.411 2.698 0.00716 **
## BusinessTravelTravel_Rarely 276.331 145.512 1.899 0.05798 .
## JobRoleHuman Resources -237.667 295.366 -0.805 0.42130
## JobRoleLaboratory Technician -900.236 195.818 -4.597 5.10e-06 ***
## JobRoleManager 4022.027 248.321 16.197 < 2e-16 ***
## JobRoleManufacturing Director -356.711 193.798 -1.841 0.06611 .
## JobRoleResearch Director 3791.686 250.255 15.151 < 2e-16 ***
## JobRoleResearch Scientist -509.365 195.868 -2.601 0.00951 **
## JobRoleSales Executive -252.517 168.092 -1.502 0.13349
## JobRoleSales Representative -775.396 241.189 -3.215 0.00137 **
## JobLevel 2779.291 96.307 28.859 < 2e-16 ***
## TotalWorkingYears 36.430 8.802 4.139 3.93e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1080 on 686 degrees of freedom
## Multiple R-squared: 0.9472, Adjusted R-squared: 0.9462
## F-statistic: 946.6 on 13 and 686 DF, p-value: < 2.2e-16-> Model 3 has Multiple R-squared ~ 0.94 and p-value < 2.2e-16. The values are almost similar to Model 1 and MOdel 2 but the same accuracy level in modelling is achieved in Model3 with less indpendent variables. Hence, best model that can be considered for estimation is Model 3.