Chapter 3

Research Methodology
 Chapter 3

CHAPTER 3
RESEARCH METHODOLOGY

3.1 INTRODUCTION

This chapter discusses the research framework and methodology of the present study.
Among several other methodological approaches, cross sectional research methodology
has been adopted. First of all, literature survey has been conducted and hypotheses have
been framed. Later on, quantitative analysis has been performed after collecting the
primary data on the criteria adheres to the same prospects.

3.2 RESEARCH PROCESS

Research process consists of a set of six steps which include problem definition,
development of an approach to the problem, research design formulation, fieldwork,
data preparation and analysis, and report preparation and presentation (Malhotra and
Dash, 2016). Similar steps have been followed to conduct the present research.

3.2.1 Problem Definition

The present study utilizes the literature survey to define the problem. Information
related to the perspective of HR professionals on integration of Artificial Intelligence
has been collected through extant literature survey.

3.2.2 Development of an Approach to the Problem

After defining the research problem, the researcher has developed an approach to the
problem using formulating research questions and hypotheses. Conceptual model has

44
 Chapter 3

been developed defining the exogenous variables to the functions of human resources
management.

3.2.3 Research Design Formulation

Research design formulation details the procedures critical for retrieving the necessary
information, to test the hypotheses of interest, determining the probable answers to the
research questions. Present study conducts exploratory research design, precisely
defines the variables of interest, and designing the survey scale.

3.2.4 Fieldwork or Data Collection

A survey questionnaire has been developed to collect the required data from the target
respondents.

3.2.5 Data Preparation and Analysis

Collection of data has been followed by filtration process. Data analysis has been
conducted on the appropriate responses.

3.2.6 Report Preparation and Presentation

Final step is to summarize the findings based on the data analysis. A report has been
prepared and presented. Inferences have been drawn based on findings.

3.3 RESEARCH DESIGN

Research design refers to the theoretical framework or blueprint of the research

conducted. It details and structures the research to obtain the information needed. It
basically defines the nature of the research and outlays process of conducting it. A
research design is comprised of defining the information needed, defining the phases
like exploratory, conclusive, descriptive/causal of the research, development of
measurement scale, pretesting the measurement scale, specifying sampling procedure

45
 Chapter 3

and sample size, developing a plan and identifying suitable techniques for the analysis
of the data collected.

Figure 3.1: Research Design by Malhotra (2008)

The first phase of present research is a combination of exploratory, conclusive and

descriptive phases. Since functions of human resource management has an important
implication on enhancing the organizational performance. Therefore, in exploratory
phase of the research, the researcher has defined the functions of human resources in
India. The researcher also tries to seek the role of artificial intelligence enhancing the
effectiveness of HRM functions in an organization. A literature survey is conducted to
gain additional insights to develop an approach to the research. Analysis of industry
reports and other secondary sources is more appropriate for the exploratory phase.
Hence a literature survey is conducted to gain additional insights so as to develop an
approach to the research. In conclusive phase, it determines, evaluates and selects the
best course of action to assist the decision maker in a given situation. The objective of
the conclusive design is to test the hypotheses and examine the relationships. The
information needed to answer the research problem is clearly defined. Furthermore,

46
 Chapter 3

sample size is large and representative to the population under study. Conclusive
research design is either descriptive or causal in nature. The present study is descriptive
in nature as it describes the functions of human resource management in IT sector in
India.

In the first phase, problem identification, development of hypotheses, and sampling

procedure and sample size has been defined. The second phase of the present research
is well organized and formal. It employs cross sectional design. At this stage,
information required is collected from the sample of the population. Therefore, the data
has been collected from human resources professional working in IT sector in India.
Since the data has been collected in single time period only, hence the present study
utilizes single cross sectional research design.

The following hypotheses have been formulated to test the research objectives.

H1: Artificial Intelligence has a significant influence on Recruitment and Selection.

H2: Artificial Intelligence has a significant influence on Training and Development.

H3: Artificial Intelligence has a significant influence on Performance Management.

H4: Artificial Intelligence has a significant influence on Task Complexity.

H5: Artificial Intelligence has a significant influence on Usefulness.

H6: Task Complexity has a significant influence on Recruitment and Selection.

H7: Task Complexity has a significant influence on Training and Development.

H8: Task Complexity has a significant influence on Performance Management.

H9: Usefulness has a significant influence on Recruitment and Selection.

H10: Usefulness has a significant influence on Training and Development.

H11: Usefulness has a significant influence on Performance Management.

H12: Task complexity moderates the relationship between Artificial Intelligence and
Recruitment and Selection.

H13: Task complexity moderates the relationship between Artificial Intelligence and
Training and Development.

47
 Chapter 3

H14: Task complexity moderates the relationship between Artificial Intelligence and
Performance Management.

H15: Usefulness moderates the relationship between Artificial Intelligence and

Recruitment and Selection.

H16: Usefulness moderates the relationship between Artificial Intelligence and

Training and Development.

H17: Usefulness moderates the relationship between Artificial Intelligence and

Performance Management.

3.4 RESEARCH INSTRUMENT

When the characteristics of a particular object are measured by assigning numbers or

symbols is known as measurement (Malhotra, 2008). The present study administered a
close ended questionnaire as research instrument to measure the characteristics and
perception of human resource professionals towards the functions they perform in an
organization. A close ended questionnaire is designed to inquire into the specific
information by providing choices and options (Malhotra and Dash, 2016). Bryman and
Bell (2015) suggest that such types of research instrument are easy to administer and
code the required information. It helps both the researcher and respondent consume less
time. Several researches have utilized the close ended questionnaire to inquire into the
performance of human resources personnel and professional in organizations. In line
with the preferences shown by the previous researcher to employ close ended
questionnaire, the present study also adopts such instrument to examine the functions of
human resource management.

The survey instrument, questionnaire has been designed into two sections. First section
consists of questions inquiring about the respondents‘ demographic characteristics such as
gender (male/female) and educational qualification (graduate/postgraduate/doctoral), and
professional characteristics like designation (Head of HR, Talent Acquisition, HR
Business Partner, Learning and Training, Compensation, and Benefits, Employee
Relation, Other), position (Lower/Middle/Senior), and experience (0-1 years, 2-5 years,
6-10 years, 10-15 years, above 16 years). Responses in this section have been recorded
under nominal scale as it requires no specific origin and order (Malhotra and Dash, 2016).

48
 Chapter 3

The second section of the questionnaire comprised of the statement pertaining to the
perception of human resource professional towards the role of HR in AI enabled human
resource functions, AI integrated recruitment and selection system of your organization,
learning and development, performance management system, task complexity, and
usefulness for HR in AI enabled HR functions. Table 3.1 depicts the measurement scale
used in present study.

The statement in the second sections of the questionnaire has been designed on five-
point Likert scale where 1 stand for strongly disagree and 5 for strongly agree. Likert
scale has been developed by Likert (1931) which is popular among the cross-sectional
studies conducted to gauge the attitude and behavior of the respondents (Bryman and
Bell, 2011). Symbols or numerical values from 1 to 5 assigned to the responses are
useful for the purpose of analysis (Malhotra and Dash, 2016).

To minimize the error and increase the scope of the scale, the researcher has suggested
the use of multiple items scale instead of single-itemized scale to determine a factor
(Spector, 1992). Keeping in mind the suggestions made by the previous researchers, the
present study also uses multiple items scale where sum total score of each item or
average score of all item forms a factor (Shiu et al., 2009).

Table 3.1: Constructs and their Respective Scale used in Study

Construct Statements
My tasks have become simple and uncomplicated with AI integration.
My task requires me to keep track of more than one thing at a time with AI
integration.
Task Complexity My task requires me to analyze a lot of information with AI integration.
My task requires that I engage in a large amount of thinking with AI
integration.
My job requires that I only do one task or activity at a time.
AI enables progress tracking by providing continuous feedback
AI monitor the performance of employees throughout the year helps in
increasing employee productivity
Performance
AI can identify and assess specific candidates for successor- leaders
Management System
AI integrated performance management system enhances employee satisfaction
AI integrated performance Management System make effective competency
mapping

49
 Chapter 3

Construct Statements
Using AI enabled HR system help me to better manage and keep track of my
task.
Using AI enabled HR system is easy to do what I want to perform in my task.

Usefulness It is easy to learn to operate AI enabled HR system

Using AI enabled HR system rarely makes me frustrated.
Using AI enabled HR system makes my work error free.
Using AI enabled HR system has been clear and understandable.
AI integration in HR functions makes my role as a HR professional more
effective.
AI integration with HR enhances the efficiency of HR functions.
Integration of HR professionals have the required skills to use AI integrated HR functions.
Artificial Intelligence AI integrated HR functions gives HR professionals more time in doing strategic
work for Organization.
It is easy to learn AI integrated HR functions by HR Professionals.
AI in HR reduces transactional work in HR Professionals.
AI helps HR Professional to compose more effective job descriptions.
AI helps HR professionals to create relevant job advertisement to attract
appropriate candidates.
AI makes background screening process faster, more accurate and more
efficient.
Recruitment and
Selection AI reduces the time during recruitment process
AI lacks human touch to attract the candidates while applying for a job.
AI creates an excellent candidate experience while hiring
AI integrated recruitment process reduces cost per hire
AI integrated recruitment enhances the performance of recruiter.
Automation of Learning and Development with AI save time for HR
professionals
AI helps HR professionals to personalizing the learning and reinforcement
processes that enhances engagement
AI helps in personalizing the learning and reinforcement processes that improve
completion rates.
AI helps HR professionals to measures learning effectiveness effectively
Training and
Development AI can create a learning program based on each employee‘s personal
preferences.
AI helps HR professionals to develop training programs accessible to learner at
any level.
AI can measure the actual outcomes of learning as it‘s an effective way of
connecting learner
AI automates many tasks and procedures in learning and development to save
time of HR professionals and reduce costs also.

50
 Chapter 3

The present study employs the scale items as developed by the previous researchers.
Moreover, since the respondents of the present study are Indians, a few new scale items
have been added and some are rephrased for better and clear understanding to the
constructs in order to reduce the influence of cross cultural and demographic variations
on the responses (Malhotra and Dash, 2016).

3.5 PRETESTING AND VALIDITY

Malhotra and Dash (2016) define pretesting as a process of improving the clarity and
comprehensiveness of the questionnaire by testing it on small sample. It can identify
and eliminate the potential problem and increases the possibility of getting suitable
responses. In pretesting, aspects such as question content, wording, sequence, from,
layout, questions difficulty, and instructions are tested. Since pretesting also requires
sample selection, it should be drawn from the same population where the actual data
has to be collected. Pretesting also include the validity assessment of the scale items.
Validity refers to the verification of the scale items by assessing whether the scale items
are describing what they actually meant for. Validity assessment consists of content
validity also known as face validity, criterion validity, and construct validity (Malhotra
and Dash, 2016). Content validity is a subjective but systematic evaluation of ensuring
the representation of scale content to the measurement task at hand. Similarly, criterion
validity ensures the performance of the scale in relation to the other variables selected
as meaningful criteria. During the pretesting, content and criterion validity has been
fulfilled. For present study in pretesting the content validity done by the experts.
Pretesting questionnaire from the experts, researcher tries to probe the uneasiness,
confusion or resistance in the respondents. Debriefing procedure has been employed to
probe any kind of correction required for questionnaire. After completing the
questionnaire, HR professionals were asked about describing the questions given in the
questionnaire. Interviewer tries to understand whether the respondent had encountered
any difficulty while responding. Based on pretesting a few questions have been
rephrased for clarity.

Pretesting has resulted into the questionnaire consists of 38 items indicating one
independent variable and one dependent variable. Two moderators also used for this
study. The questionnaire with 38 items was found to be sufficient for the getting
responses to understand the perception of HR professionals towards use AI in HR

51
 Chapter 3

functions. After completing the pretesting, the questionnaire was subjected to pilot
testing.

3.6 PILOT TESTING

Before moving to the actual data collection, pilot survey is important to assess the
response pattern from the participants. Moreover, pilot surveys are found to be helpful
in identifying the one-dimensionality of the scale by employing exploratory factor
analysis (discussed in Chapter 4). Pilot survey are initial testing of the responses to be
generated using questionnaire, these are generally conducted on the smaller sample size
than the large sample for actual data collection. One of the popular rules of thumb
suggested by the Hair et al. (2010) and Cooper and Schindler (1998) are that the
researcher can use 25 to 100 responses to pilot testing. Hence the present study uses 100
responses for pilot testing. For present research, the sample for the pilot survey includes
the HR professional working in IT sector in India where AI is integrated with HR
functions. Table (3.2) depict the result of pilot test using explorotary factor analysis and
Table (3.3) shows the result of the reliability test to acess the consistency of the items.

Table 3.2: Rotated Component Matrix

Component
1 2 3 4 5 6
AIRS7 .773
AIRS3 .772
AIRS5 .769
AIRS4 .766
AIRS2 .765
AIRS6 .761
AIRS1 .759
AIRS8 .757
AITD3 .745
AITD5 .743
AITD7 .737
AITD4 .728
AITD8 .721
AITD2 .689
AITD6 .683

52
 Chapter 3

Component
1 2 3 4 5 6
AITD1 .675
UHRAI3 .735
UHRAI6 .731
UHRAI4 .721
UHRAI1 .713
UHRAI2 .710
UHRAI5 .668
RHRAI2 .756
RHRAI3 .714
RHRAI4 .687
RHRAI5 .681
RHRAI1 .676
RHRAI6 .667
TCAI3 .793
TCAI4 .781
TCAI5 .753
TCAI1 .718
TCAI2 .657
AIPMS3 .724
AIPMS2 .696
AIPMS1 .651
AIPMS4 .643
AIPMS5 .633
Extraction Method: Principal Component Analysis.
Rotation Method: Varimax with Kaiser Normalization.
a. Rotation converged in 6 iterations.

Table 3.3: Reliability Analysis

S. No. Variables Alpha

1. RHRAI .723
2. AIRS .789
3. AILD .711
4. AIPMS .825
5 TCAI .866
6. UHRAI .768
Note: RHRAI = Role of artificial intelligence; AIRS = Recruitment and
Selection; AITD = Training and Development; AIPMS = Performance
Management: TCAI = Task Complexity; UHRAI = Usefulness

53
 Chapter 3

3.7 SAMPLING PROCEDURE

To obtain the information about the characteristics or parameter of the population,

sampling procedure involves the selection of the sample which is the true representative
of the same population (Malhotra and Dash, 2016). In sampling, unlike census where
data is collected from each element of the population, the required data is collected
from the sample selected. Census is a costly affair and also makes it difficult to the
researcher to analyze the data as being the dynamic nature of the population. Hence the
present study adopts the sampling approach to collect the required data. Malhotra and
Dash (2016) suggest the following steps required in sampling procedure. Sampling
procedure includes steps depicted in Figure 3.2.

Figure 3.2: Sampling Procedure by Malhotra (2008)

3.7.1 Target Population

The target population is comprised of the collection of sampling elements or objects

that possess the information sought by the researcher about which the inferences are to
be drawn (Malhotra and Dash, 2016). The target population is defined in terms of
sampling elements, sampling units, extent and time (Malhotra and Dash, 2016). The
sampling element is the respondents whom the information is to be sought. Appropriate
selection of sampling element is necessary in order to reduce the unbiased responses
and increase the relevance. Several previous researches have examined the

54
 Chapter 3

effectiveness of the HRM functions considered the HR professionals as their sampling

element (Agarwal, 2022; Anger et al., 2021; Guliyeva et al., 2020).

Present research considers male and female HR professionals above 18 years of age
working in IT sector as sampling elements. Sampling unit for the current study is the
companies in IT sector under PAN India. NASSCOM 2021-22 reports the IT sector has
recorded the highest growth in the last decade with 15.5 percent jump and set to become
$227 billion industry in India. Moreover, India innovation index 2020 of NITI Aayog
suggests that Delhi is the top performer in innovation among the union territories in
India. Delhi has scored 46.60 points followed by Chandigarh scoring 38.8 points.
Extent of the current sampling procedure is the geographical boundary i.e., India for the
present research. The time period for the survey was July 2021 to December 2021.

3.7.2 Sampling Frame

Sampling frame consists of a list or a set of directions for identifying the target
population (Malhotra, 2008). The examples of sampling frame include telephone book,
association directory, mailing list purchased from an organization, city directory or a
map. There are certain cases where the list cannot be compiled. In such cases some
directions like random digit dialing number in telephone surveys may be identified for
the target population. However, the present research uses convenient sampling
(discussed in following section), wherein the researcher employs the sampling method
at its convenience and sampling frame in such case may not be required.

3.7.3 Sample Technique

Once the sample is selected from a given population, the researcher, based on the nature of
the research has to decide whether the data collection and analysis to be done
simultaneously. There are two approaches to decide the same. Traditional approach
suggests that the all the data has to be collected before the data analysis begins unlike
Bayesian approach where with the addition of each sample element the analysis is revised.
Since the traditional approach simple and smooth, it is highly popular among behavioral
researches. Hence the present research also employs tradition approach of sampling.

55
 Chapter 3

After deciding upon the approach to the sampling, the challenge face by the researchers
is to determine the method of selecting the elements from the sample. Hence the most
important decision is about the choice of the sampling between probability and non-
probability technique. In probability sampling, the elements in the sampling frame have
equal chance of being selected. However non-probability sampling relies upon the
personal judgment of the researcher rather than chance to select an element (Malhotra,
2008). Researcher in non-probability sampling consciously decides about the elements
to be included in the sample. Although probability sampling is more scientific and
superior to the non-probability sampling, but availability of reliable sampling frame is
another limitation with the researcher (Dörnyei and Taguchi, 2009). Hence, in such
scenario to get the sample which is reasonable and true representative of the population,
it is about the availability of the resources to the researcher to use the different methods
of sampling. For present research, non-availability of the reliable source of sampling
frame leads researcher to opt for Convenience sampling or researcher-controlled
sampling approach. Convenience sampling is common approach in behavioral research
as it helps researcher to reach large number of respondents in cost effective manner
(Malhotra and Dash, 2011; Gravetter and Forzano, 2003).

3.7.4 Sample Size

Sample size refers to the number of elements to be included in the sample. Malhotra
and Dash (2016) state that determining the sample size is complex and requires few
qualitative and quantitative considerations. Qualitative considerations include nature of
research, number of variables, nature of analysis, sample size used in similar studies,
resource constraint etc. Quantitative considerations include statistical approach where
the precision level for the sample is pre specified. Statistical approach defines the
confidence interval around the means and proportions (Malhotra, 2008). Since the
statistical approach is viable method for probability sampling, present study found its
basis of determining sample size in qualitative considerations. Researcher of the present
study opts for nature of analysis and sample size used in previous studies as a basis of
determining the sample size. As the present study has employed multivariate data
analysis techniques (SEM) which is highly sensitive to the size of the sample, it is
important for the researcher to collect sufficient number of responses. Less number of
responses generates unreliable and unstable results in SEM (Tabachnick et al., 2001).

56
 Chapter 3

Hair et al. (2019) suggests proposed two principles i.e., minimum number of respondents
and subject-item ratio for determining the sample size based on the SEM methodology
employed in the analysis. The first principle states that the minimum number of the
responses should be decided for the analysis (Garson, 2008). MacCallum et al. (1999)
recommends the thumb rule of 100 respondents is sufficient for the analysis using SEM.
There are other previous researchers who have suggested the 100 responses are good
number for the analysis (Hatcher, 1994). He also states that the number of responses
could also be 5 times the number of items formulated in the scale. Comrey and Lee
(1992) proposed the thumb rule of 500 responses for the analysis using SEM. Another
principle of subject item ratio is also popular employed while determining the sample
size. Bryant and Yarnold (1995) recommend the 5:1 subject item ratio can be appropriate
sample size. Here 5 times the responses of all scale items can cumulatively represent the
size of the sample. Other researchers have recommended the greater number of responses
such 10:1 or 20:1 for determining the sample size (Hogarty et al., 2005; Velicer and Fava,
1998). The present adopts the 10:1 subject item ration method. The scale items in the
present research are 38 which make the total number of responses equivalent to 380 (Hair
et al., 2019). However, the researcher has targeted to collect 500 responses so that after
data refinement, at least 380 appropriate responses are received.

3.8 FINAL DATA COLLECTION

After finalizing the method of selecting sampling elements, the next step is to collect
the final data for the study. Since the objective of applying the non-probability sampling
is reach out to 500 respondents and the researcher is situated in Delhi NCR and
collecting data form respondents residing in distant locations like Bengaluru, Chennai
and Hyderabad was a challenge. Hence the online survey has been considered a suitable
option. For collecting the data online, Google form has been prepared (Chakraborty,
2019; Dhir and Dutta, 2020). For the purpose, link of Google form has been distributed
through emails, WhatsApp and other social media platforms. As previously mentioned,
the data collection drive was conducted during July 2021 to December 2021. Total of
487 responses has been generated. Out of total responses, Convenience sampling has
resulted in some questionnaires from non-qualified respondents which had to be
deleted. Only 418 responses deemed fit and appropriate for further analysis.

57
 Chapter 3

3.9 DATA SCREENING

Once the data is collected, researcher is required to the check the quality of the data in
completed questionnaire (Malhotra and Dash, 2016). It requires screening the data in
order to prepare for the analysis. Data screening is processing the data collected to make
fit for the analysis. It involves the procedure like checking of completed questionnaire,
editing, coding, transcribing, and data cleaning to ensure the uniformity and consistency
of the data. Later, the responses received in form the completed questionnaire are
transformed into meaningful code as a prerequisite to data analysis. Table 3.4 highlights
the name of demographic variables and their respective code.

Table 3.4: Coding and their Respective Variables

[Link]. Variable Code

1. Gender Male = 1; Female = 2

2. Designation Head of HR =1, Talent Acquisition = 2, HR Business Partner = 3,

Learning and Training = 4, Compensation, and Benefits = 5,
Employee Relation = 6, Other = 7

3. Current Position Lower = 1; Middle = 2; Senior = 3;

4. Length of Service 0-1 years; 2-5 years; 6-10 years; 10-15 years; Above 16 years;

5. Educational Qualification Graduate = 1; Postgraduate = 2; Doctoral = 3;

Moreover, responses for artificial intelligence, task complexity, usefulness, recruitment

and selection, learning and development and performance management has received on
five-point Likert scale. Likert scale used in present study denotes 1 as strongly disagree,
2 as disagree, 3 as neither agree nor disagree, 4 as agree and 5 as strongly agree.

3.9.1 Missing Data

Missing data refers to the observation or values recorded in the questionnaire are
unknown due to the respondent has either provided ambiguous answer or it is not
properly recorded. It also represents the non-response which arises because the
respondent has refused responding. Hair et al. (2019) argue that missing data can lead to
reduced sample size and non-random missing data processing can lead to inaccurate

58
 Chapter 3

results. Malhotra and Dash (2016) suggest the treatment of the missing observation
depends upon its proportion on the overall responses. If the proportion of missing data
is less than 10 percent, it can be treated by substituting it with a neutral value or with an
imputed response. Otherwise, in case where the proportion of missing data is greater
than 10 percent, individual case ought to be omitted. Since proportion of the missing
data in the present is less than 10 percent, researcher has substituted it with the neutral
response. For itemized scale, the neutral response is 3. Computing imputed response for
the missing observation, considerable efforts are required and can cause some serious
biases (Malhotra, 2008). Therefore, some sophisticated statistical procedures are
required to compute the imputed response. Due to this reason, present study inputs
neutral values to substitute missing response.

3.9.2 Outliers

Outliers are the extreme values which are unique and distinct from being normal (Hair
et al., 2019). These extreme values can be large as well small. Outliers are the values
which are not similar to the values observed by the researcher from the same respondent
(Dörnyei and Taguchi, 2009). Hair et al. (2019) defines the outliers as the values which
are different from the rest of the observations. The treatment of the outliers is necessary
as it increases the diversity of the data being processed for the analysis. For present
study, researcher has used box plot to identify the extreme cases. Resultantly six cases
were identified which contain the extreme responses. Such cases were not deemed fit
for the analysis, hence deleted.

3.9.3 Normality Analysis

Assessment of normality is one of the important assumptions to process the data for the
actual analysis. The benchmark for the statistical analysis is the normal distribution of
the data. The shape of the distribution of the responses generated over the metric
variables of the study should correspond to the normal distribution (Hair et al., 2019).
Larger variation from the normal distribution could lead to invalid results. Hence it

59
 Chapter 3

becomes important to assess the normality of the data. Hair et al. (2016) argue that the
presence of the non-normality is based either on the shape of the offending distribution
or size of the sample. In former case, the departure of the distribution from the normal
shape can be characterized as kurtosis or skewness. Identification of such non-normal
shapes is crucial. Kurtosis is peakedness or flatness in the distribution which is termed
as leptokurtic and platykurtic respectively. Skewed distribution is unbalanced
distribution and responses are consolidated at any of the sides of the graph. The
statistical measure to assess the skewness and kurtosis of the normal distribution gives
the value zero (Hair et al., 2019). The effect of the sample size on the normality of the
distribution should also be considered by the researcher. Large sample size is
detrimental to the non-normality of the distribution. Hair et al. (2019) state that the
small sample less than 50 leads to departure from the normal distribution and gives
biased results. However, sample size greater than 200 makes such departure negligible.
Graphical analysis and statistical analysis are the main approaches to diagnose the
normality of the data. The current study has fulfilled the assumption of normality by
employing the statistical method. Statistical method includes the assessment of
normality by confirming the absence of skewness and kurtosis in the distribution.
Moreover, the researcher has also taken the large data size so as to deter the effect of
non-normality on the results.

3.10 STATISTICAL TESTS FOR THE DATA ANALYSIS

The present study examines the impact of artificial intelligence on the effectiveness of
the functions of HRM. The conceptual model of the present study defines the
recruitment and selection, training and development and performance management
system as dependent variables, task complexity and usefulness as moderating variables.
Moreover, role of artificial intelligence is the independent variable. Such a complex
model requires multivariate analysis techniques for estimating the results. For the
purpose, Structural equation modeling has been applying to examine the direct and
mediating effects of the conceptual model. For analyzing the moderating effects,
ordinary least square using Hayes‘ PROCESS has been employed.

60
 Chapter 3

3.10.1 Structural Equation Modeling (SEM)

Malhotra and Dash (2016) states that the structural equation modeling estimates the
interrelationships among the multiple variables of the study. It examines the structure of
interrelationship defined in the terms of series of equations similar to the multiple
regression analysis (Hair et al., 2019). The interrelationships are examined among the
constructs which can be dependent, independent, moderating variables. The constructs
in the structure are not directly observed hence represented by the number of variables.
Moreover, SEM is the foundation of interdependence and dependence technique which
is familiar to the multivariate techniques comprised of factor analysis and regression
analysis.

A model is a representation of theory which is a systematic set of relationships

providing a consistent and comprehensive explanation to the phenomenon (Hair et al.,
2019). Since theory is not exclusive area of academics, it draws its formation from the
real-world experiences. The conventional model in SEM comprised of measurement
model and structural model. Measurement model depicts the formation of measured
variables that represents a construct. Structural model highlights the association
among the constructs. Below are the four characteristics of SEM stated by the Hair et
al. (2019):

1. SEM can examine the multiple relationships between dependent independent and
moderating constructs.

2. SEM helps estimating the latent constructs which are not directly observed from
the respondents and accounts for measurement error.

3. SEM draws upon the theoretical model which defines the entire relationship

4. SEM also helps checking for over-identified, under-identified, and just identified
model. Model identification signifies that the variables explained by a unique set
of variables do include all the possible relationships.

The present study employs following steps for the conducting the structural equation
analysis.

61
 Chapter 3

Figure 3.3: SEM Flowchart by Malhotra and Dash (2016)

The present study has developed measurement model to validate the scale using
confirmatory factor analysis and structural model to test the path estimates. Present
study has utilized the SEM for estimating the direct and indirect effects.

3.10.2 Ordinary Least Square using Hayes’ Process

One of the objectives of the study is to examine the influence of moderating effect of
artificial intelligence and usefulness/task complexity on each of the dependent variables
named as recruitment and selection, learning and development and performance
management system. Moderation analysis has been conducted using Hayes‘ PROCESS
which employs ordinary least square to estimate the conditional path between
dependent and independent variables. Hair et al. (2019) defines the moderation as
inclusion of third variable which changes the strength of the relationship between two
constructs. The moderating variable can metric and non-metric. Non metric moderating
variables are categorical in nature such gender and educational qualification etc. The
metric moderator is continuous in nature. The present study uses task complexity and
usefulness as metric moderator for the analysis.

3.11 SUMMARY

For this study, the methodological approach that was outlined above was utilised. In
Chapter 4, both the result and interpretations of the data are presented.

62