The Impacts of Requirements Relationships Knowledge on Requirements Quality and Software Development Project Success

Abstract

Requirements quality is one of the factors that may determine the success or failure of a software project. Thus, maintaining requirements quality is important but also a challenge as an individual requirement does not stand alone and they are related to one another in several ways. The problem may become more challenging as the requirements and their interrelationships are not static and will continually change. However, current research largely focusing on the assessment of the impact of requirements quality on success. There is lack of research assessing the impact of the interrelationships between requirements on success. Therefore, this research aims to investigate how the interrelationships between requirements impact requirements quality as well as the success of software development project. An empirical study to examine further the impacts was conducted from the perspective of business analyst. Using Structural Equation Modelling (SEM) and especially Partial Least Square (PLS), we found that there are significant impacts of requirements relationships towards requirements quality as well as success. The outcome from this research can be used as a guide to working with requirements relationships, knowledge useful for business analysts and research community.

Keywords

• eol>Requirements Relationships Knowledge
• Requirements Quality
• Software Development Project Success
• Partial Least Square (PLS)
• CEUR-WS

-

1. Introduction relationships between requirements is defined as requirements relationships knowledge (RRK). RRK is concerns Requirements quality is defined as a set of requirements on how requirements are related to one another and other or software requirements specification (SRS) that having artefacts during the software development project. Acall the good characteristics that listed as proposed by cordingly, requirements relationships knowledge may IEEE-830-1998 recommended practices for SRS [1]. Ac- provide guide in organising and structuring the requirecording to the practice; attribute of requirements quality ments documentation and specification. Karlson et al. [3] are including correct, unambiguous, complete, consistent, indicated that one of the main contributions of requireranked for importance and/ or stability, verifiable, modi- ments relationships knowledge is in the bundling strucifable and traceable. Thus, an SRS developed for a partic- ture of requirements. A good structure and organised ular software project should fulfill all the characteristics requirements specification can facilitate better managelisted to ensure the requirements quality. Consequently, ment of requirements, whether it is done manually or by to produce quality requirements specification; the com- any automatic tool. It will also provide a good basis for prehensive understanding of requirements is needed. any manipulation and maintenance activities for the later

Moreover, to fulfil the necessity of comprehensively phases. This will increase the possibility of achieving understanding requirements, it is important to acknowl- project success. Diev [4] ascertained that requirements edge how each requirement is related to one another. structuring is an essential activities in requirements enKnowledge on how each requirement is related to one an- gineering as requirements structure and representation other may assist stakeholders to make informed decision will directly impact requirements development process in accomplishing many things that involve in managing and the requirements quality. The importance of requirerequirements [2]. In this paper, the information of the ments quality is also advocated by agile practitioners [5, 6].

In relation to this, previous researchers asserted that

requirements quality especially Software Requirements Specification quality has strong impact on the success or failure of a software development project [ 7, 8 ]. They have made thorough investigation into how requirements quality impacts project success. There are also some researchers that suggest the contribution of RRK [3, 4, 9]. RE or badly performed RE activities may result in incorHowever, there are lacks of study that examine how RRK rect and incomplete requirements, besides the possibilimpacts requirements quality as well as software develop- ity of high rate of changing in the requirements, which ment project success. Therefore, this research will extend could be the reason for the software project to be chalthe previous works to examine further the three main lenged. Badly performed RE process has been claimed issues: 1) the impacts of RRK on requirements quality; 2) as positively associated with software failure [16, 17, 18]. the impacts of requirements quality on project success; 3) Therefore, by improving RE practices, there would be the impacts of RRK on project success. This research has an economic as well as software quality payof [16, 17]. been conducted empirically using survey method and the Clearly, literature indicates the importance of requireanalysis of the data has been performed using Structural ments quality and requirements engineering as the critiEquation Modelling (SEM) and specifically Partial Least cal success factor of a particular software development Square (PLS). project. In relation to this, Verner et al. [19] argue that

The rest of this paper will be organized into 4 sections. the most important correlation in achieving project sucFirst, the research context and research model is dis- cess is to have good requirements and to manage those cussed in Section 2. Second, the research method, which requirements efectively. However, to get the correlations mainly concerns on the development and the validation of both factors, we need to know how requirements are of the requirements relationships instrumentation de- related to one another which a concern of RRK. sign, will be discussed in Section 3. Next, section 4 of this Although requirements relationships in any software paper will present the discussion of the result. Finally, development project are not problematic, they would afsection 5 will present the concluding remark, including fect other aspects of software development project and the future work in both research and practice. the project as a whole. RRK is asserted as essential when making decisions in the subsequence phases of any software development project including designing [20], re2. Research Context quirements prioritisation [21] and testing [22]. In addition, failure to consider RRK during requirements activities is argued could lead to costly mistakes [ 23 ]. Hence, RRK needs to be carefully identified, analysed, and managed to avoid any ripple efects.

Moreover, the success of requirements engineering in producing requirements quality as one of the success factors of software project has been discussed in many studies (e.g. [ 16, 17, 23 ]) but the studies that particularly discuss how RRK impacts project success are limited.

Hence, the questions to be asked are, 1) is this knowledge (RRK) really significant in software development project? 2) If yes, how can RRK impact requirements quality as well as the success of software development project? 3) Other than that, in what way can this knowledge be fully utilised for that purpose? In order to answer these questions, this paper aims to discuss these issues further and extend the literature on the interrelationships between RRK, requirements quality and the related issues that have impacts on project success. The related research model are proposed and illustrated in Figure 1. The model was developed based on software project success factors that are related to RRK as discussed in literature. However, this paper will be focusing only on a part of the model in which consists of the three constructs: 1) RRK, 2) Requirements Quality, 3) success. Thus, the related hypotheses are as follows:

H1: RRK has significant impact on requirements qualThe success of a software development project (SDP) is a concern for any related stakeholders. Success in SDP is described based on several criteria including: 1) quality of product [ 10, 11, 12 ], 2) Timeline of the delivery (schedule) [ 11, 12, 13, 14 ], 3) Cost [11,12] , 4) Satisfaction of stakeholder [11], 5) met requirements [11], 6) met business objective [11] 7) met scope [12] and 8) learning [14].

Other than that, study in software project management will also involve the factors that may impact the success.

The success of a software development project (SDP) is a concern for any related stakeholders. Success in SDP is described based on several criteria including: 1) quality of product [ 10, 11, 12 ], 2) Timeline of the delivery (schedule) [ 11, 12, 13, 14 ], 3) Cost [11,12] , 4) Satisfaction of stakeholder [11], 5) met requirements [11], 6) met business objective [11] 7) met scope [12] and 8) learning [14]. Other than that, study in software project management will also involve the factors that may impact the success.The results of both studies have shown the significant impacts of requirements quality on success. The later researchers ranked clear requirements and specification as the top factor among 26 critical success factors of software development project [ 15 ]. The ranking supports both previously bodies of knowledge in the assertion of requirements quality as the significant factor for software project success.

Accordingly, quality in requirements specification will always depends on how requirements are determined in the process of requirements determination, which is known as Requirements Engineering (RE). Not enough ity.

H2: Requirements quality has significant impact on the success of software development project. H3: Requirements relationships knowledge (RRK) has

The overview of the initial indicators for every construct in the model is shown in figure 2.

quirements in their software development project were chosen whereas those not were excluded from the samFigure 2: Initial Indicators of the Model ple. About 173 business analyst and related stakeholders were recruited. The inclusion criteria were including the respondents were stakeholders involved in the mansignificant impact on requirements quality as well as on agement of requirements in their software development the success of a software development project. project. Approximately, 60 percent of the participants

Accordingly, to validate the hypotheses, empirical were business analysts and system analysts. Most of the research analysis using Structural Equation Modelling respondents are practitioners in Malaysia Industry and (SEM) was performed. This paper will continue to dis- only 10-20 percent of the respondents are from Australia. cuss about the research method used to examine the The findings show that most of the respondents are from interrelationships between RRK, requirements quality medium and large organization (refer to Table 1) and and success of Software development project (SDP) in Australian Bureau of statistics classification of business the next section. framework (ofice of small business, 1999). In addition, in Table 2, the findings show that the industry domain of most of the organization is from Information Technology 3. Methodology and Telecommunication (41 percent) and, infrastructure and Government (30 percent). Moreover, the respondents 3.1. Participant largely have about 2-5 years (37 percent), and about 6-9 In this study, the sample was chosen using non- years (25 percent) experience in requirements writing probability sampling specifically purposive sampling. In which represent approximately 62 percent of all the rethis regard, any stakeholders involved in managing re- spondents (Table 4). Table 4 rather than exploratory. Thus, SEM is more appropriExperience in Requirements Writing ate for theory testing than theory development. SEM is a generic and powerful multivariate analysis technique

Item Frequency Percentage that includes specialised versions of several other analyWithin one Year 32 19 sis approaches as special cases. SEM is not intended for 2-5 Years 63 37 a single statistical technique but it is a family of related 6-9 Years 43 25 procedures [26]. Other related terms used are Covariance 10-15 Years 23 13 Structure Analysis, Covariance Structure Modelling, and More than 15 years 72 41 Analysis of Covariance Structures.

Moreover, SEM can be categorized into two approaches, which are: 1) covariance-based approach, 3.2. Data Collection which is related to tools such as EQS and AMOS; and 2) There were approximately 380 self-administered ques- variance-based approach, which is related to PLS. Theretionnaires used for collecting the data from the respon- fore, PLS was chosen to be used in this research. Partial dents. Several methods of questionnaire distribution Least Square (PLS) was chosen because of the following were employed: 1) a number of questionnaires were reasons: 1) research on requirements relationships is relmailed to the respondents; 2) a number of questionnaires atively new; and 2) there is no measurement model that were emailed (on-line survey); 3) a number of question- is already available. PLS is asserted as a suitable technaires were completed using drop-of survey method. A nique to be used when the phenomenon to be examined total of 210 questionnaires were received but only about is relatively new [ 27, 28 ]. Hence, the assessment of the 173 questionnaires were usable for analysis. This trans- goodness of measure of these constructs in terms of their lates to about 55.3 ’percent response rate and only 45.5 validity and reliability within the research framework percent were considered efective response rate. will be discussed in the next section.

Accordingly, missing values or data are asserted as part of almost all research [ 24 ]. One of the ways is to 3.4. Measure and Goodness of measure omit the subjects that have missing data. If there are missing data at about more than 20 percent from the A questionnaire using five-point Likert scale was deitems in a questionnaire, the subjects related are advised signed to collect data for each construct of the research to be deleted from the analysis [25]. Hence, in this study model. Some of the instruments in the questionnaire we used the usable questionnaire only in the analysis. were newly developed whereas most of the questions The technique used is also known as Listwise. Listwise were designed based on the theory from literature and is the technique where subjects are discarded from the other empirical studies. Additionally, some parts of the analysis because of there are some questions unanswered instruments were adapted from previous literature. The in the survey. Even though, this technique will decrease ifnal constructs of the model are illustrated in Figure 3: the subjects for the analysis but this technique is used to ensure that the analysis will be done with complete data for all the subjects. 3.5. Goodness of Measure

3.3. Structural Equation Modeling (SEM) and Partial Least Square (PLS)

This study is a part of a research that investigated the impacts of requirements relationships on the other factors and tasks in any software development project that possibly will also impact success. The factors and tasks may have direct and indirect relationship; they might impact one another and thus the success or failure of a particular software development project as a whole. Hence, Structural Equation Modelling (SEM) was used to validate and examine the interrelationships and the impacts that they have to one another. SEM is a statistical technique for the validation and estimation of causal relationships using a mix of qualitative causal assumption and statistical data. This method is usually used more for confirmatory

In this study, two main criteria have been utilised for

evaluating goodness of measures, which are validity and reliability. The combination of both is essential to assure the quality of a research [29]. Validity is about how well a developed instrument measures the particular concept that is intended to be measured [30]. On the other hand, Trochim and Donnelly [29] also indicated that reliability refers to repeatability or consistency. A measure is considered reliable if it gives the same result over and over again. The validity and reliability measures of this research model are discussed in the next section.

3.6. Construct Validity

Construct validity can be described as the degree to which interferences can legitimately be made from the operational constructs in a particular study to the theoretical

Construct Success Requirements Q42a Relationships (RRK1) Knowledge Requirements Q14 Quality (RQ2)

Q24 (RQ3) Q13 (RQ1) Item Description Q5 (SC1) The outcome of the project meets the business goal.[11] Q6 (SC2) The outcome of the project meets all the specified

requirements.[11] Q7 (SC3) The overall quality of the developed application / product

is high.[ 10,11 ] Q10 (SC4) The project is completed within scope.[12] Q11 (SC5) The requirements-related tasks (e.g. requirements specification, requirements management) have been completed successfully in the project.[ 11,12,13 ] The relationships between requirements that exist between the components are considered when deciding to implement the solution. [2] Q42b The relationships between requirements that exist (RRK2) between the components are considered when planning the schedule for the design/development team to complete the task. [ 20, 23 ] Q35 Before implementing a change to a particular (RRK3) requirement, any possible impact it will cause to other requirements will be considered. [39] Requirements are typically grouped according to similar functionality / business area. [ 1, 3, 4, 7 ] The requirements specified in the requirements document are easy to be located whenever needed. [1] There is a specific structure/arrangement to follow when specifying requirements in the requirements document.

[ 1,3, 4, 7 ] constructs on which those operational constructs are based on [31]. Sekaran and Bougie [30] ascertained that construct validity can be used as a confirmation on how well the results obtained from the use of the measure fit the theories around which the test is developed. Thus, convergent and discriminant validity were used to examine how the instrument fits the concept as theorised. Initially, the respective value of loadings and cross loadings in Table 6 were examined to assess whether there were any problems with any particular items. A cut-of value for loadings at 0.5 was considered as significant [25]. If there were any items with a loading of higher than 0.5 on two or more factors, then they were deemed to be having significant cross loadings. Table 5 shows that all the items that measured a particular construct would load highly on the construct and would have lower loadings values on other constructs therefore confirming construct validity.

3.7. Convergent Validity

Accordingly, the test for validity was continued with the

convergent validity. This validity test is concerned about the degree to which multiple items are in agreement to measure the same concept. Factor loadings, composite reliability, and average variance extracted (AVE) were

Item

RR1 RR2 RQ1 RQ2 RQ3 SC1 SC2 SC3 SC4 SC5

RRK used to measure the convergent validity. This practice was proposed by Hair et al. [245. The convergent validity test findings showed that the factor loadings for all items exceeded the recommended value of 0.5 [25]. Next, composite reliability values illustrated in Table 6 present the degree to which the construct indicators indicated the latent, ranged from 0.793 to 0.925. The value is apparently exceeded the recommended value of 0.7 [25]. Finally, the average variance extracted assessed the variance captured by the indicators relative to measurement error. The value should be greater than 0.5 to justify the use of the construct [30]. As illustrated in Table 6, the AVE was in the range of 0.556 to 0.861.

3.8. Discriminant Validity

Then, the study was continued to validate the discriminant validity. Discriminant validity is concerned about

0.561 Another issue in the area of survey research is common 0.185 0.556 method variance. Considering the self-reported nature of the data used, there was a possibility for this issue to happen. Hence, Harman one factor test was performed the degree to which items diferentiate among constructs to determine the extent of this issue. Accordingly, Podwhere they illustrate the measures that theoretically sakof and Organ [36] indicated that common method should not be related are in reality not related. This bias is problematic if a single latent factor would account validity test was assessed by exploring the correlations for the majority of the explained variance. The unrotated between measures of potentially overlapping constructs. factor analysis illustrated that the rfist factor accounted The items should have the highest loading value on their for only 22.5 percent of the total variance, consequently own constructs in the model, and the average variance ascertained that the common method bias was not a serishared between every construct and its measures should ous issue in this study. be more than the variance shared between the construct Finally, the analysis is continued with the path analand other constructs [31]. Table 8 illustrates that the ysis to test all the hypotheses generated in this study. squared correlation for each construct is less than the Table 10 presents the results. The result of the analysis average variance extracted by the indicators measuring shows that the three hypotheses: H1, H2 and H3 were the construct to indicate the adequate discriminant supported. The results implied that there are significant validity. As a consequence, the measurement model interrelationships between requirements relationships has demonstrated adequate convergent validity and knowledge, requirements quality and success of software discriminant validity. development project. In the analysis, the path coeficient value for RRK->RQ is 0.270 whereas the path coeficient for RQ-> Success is 0.306. Both coeficient values are in the ranges of (0.20-0.30) that have been asserted as 3.9. Reliability Analysis acceptable [ 27 ]. Hence, it can be concluded that there Reliability is concerned about the quality of measure- are significant relationships exist between the three conment. Reliability in a research is the degree to which structs. a measurement procedure produces the same answer each time the measurement procedure is carried out [33]. Table 10 One of the general classes of reliability is the internal Path Coeficient and Hypothesis Testing consistency reliability that is utilised to measure the Hypothesis Relationship Coeficient t value consistency of result across items within a test [29]. In relation to this, Cronbach’s alpha coeficient was used to HH12 RQR-R>KS-u>RccQess 00..320760 32..389689 YYeess examine the reliability of the inter item consistency of H3 RRK->Success 3.465 Yes the measurement items. The summarisation of loadings and alpha values are illustrated in Table 9. Based on the ifndings in Table 9, all the alpha values are above 0.6, Moreover, mediation efect analysis has also been

Support

which are conforming to what have been suggested by Nunnaly and Berstein [ 34 ]. Consequently, the composite reliability values also ranged from 0.793-0.925 (refer table 6). Composite reliability values are another method similar to Cronbach’s alpha for internal consistency reliability estimate where a composite reliability value of 0.7 or more is considered acceptable [35]. Therefore, it can be concluded that the measurements used in this study were reliable. conducted. The finding reports that, there is exists mediator relationships between the three constructs. Figure 4 illustrates the analysis which represents the initial coeficient for the three constructs. There are several criteria that need to be fulfilled before any mediation efect analysis can be performed. First, the predictor (RRK) has significant impact on the mediator requirements quality (RQ) (later noted as a); second, the mediator (RQ) has significant impact on the criterion variable Success (b); and third, the predictor (RRK) has significant impact on the criterion variable in the absence of the mediators’ impact (c). Therefore, to establish the mediating efect, the indirect efect of a x b (see figure 4) has to be significant. In this regard, the z statistic is applied [3], specifically the value is significant at p <0.05. If the z value exceeds 1.96 (p <0.05), then the hypothesis H3 can be accepted where there is an indirect impact of RRK through requirements quality on the success of software development project. The z value is defined as the following:

× = √︀2 × 2 + 2 × 2 + 2 × 2 (1)

As illustrated in figure 4, there is a significant impact

of RRK on requirements quality (0.271, p<0.05) as well as requirements quality on success (0. 387, p<0.05). Conse- Therefore, it can be concluded that the relationships quently, there is also a significant direct impact of RRK between the three constructs is significant and the three on the success of software development project (0.169, hypotheses are also confirmed by the mediation efects P<0.05); thus, requirements quality is established as a that exist among them. partial mediator. This mediating efect of requirements quality in this study is confirmed by z statistic [38]: 4. Result and Discussion were reliable. Accordingly, the findings of this paper con- Finally, the phenomena of the interrelationships beifrmed and supported the direct significant impacts of tween RRK->Requirements quality ->Success (H3) has RRK on requirements quality (H1); the direct significant been proved by the findings. The interrelationships beimpact of requirements quality on the success of software tween them are also found to be the strongest link that development project (H2); and consequently supported existed in this study. hypothesis (H3) that proposed the indirect significant impacts of RRK on the success of software development project. 5. Conclusion

Firstly, RRK has significant impact on requirements quality in which inline with what have been ascertained Therefore, the findings have confirmed the three hypothein the literature (e.g. [ 23 ]) in which supporting H1. The ses listed in this study. As requirements relationships requirements relationships knowledge provides guide knowledge has significant impact on requirements qualon how a set of requirements can be structured and or- ity (H1); and requirements quality has direct significant ganized in requirements specification document. The impacts on success (H2); it can be concluded that rerequirements documentation that is properly organized quirements relationships knowledge is another signifiand well structured can contribute to the good quality cant factor that will impact requirements quality as well of requirements [ 4, 23 ]. According to the analysis of the as project success (H3). Accordingly, the findings also result, the main characteristics of requirements quality confirmed the significant impacts of RRK on the software that related to RRK are: 1) Requirements are typically project success. In the future, this quantitatively finding grouped according to similar functionality/business area; of this study will be continued with a qualitative study 2) The requirements specified in the requirements doc- in investigating further how RRK impacts requirements ument are able to be located whenever needed; and 3) quality and other related factors on the software project There is a specific structure / arrangement to follow when success from the business analyst perspectives. specifying requirements in the requirements document.

Characteristics of items 1 and 3 confirmed the important Acknowledgments of RRK in structuring requirements in an SRS. Both items then may support the characteristics of items 2. When The authors fully acknowledged Ministry of Higher Eduthe requirements can be located whenever needed, an cation (MOHE) and Universiti Tun Hussein Onn Malaysia SRS can be indicated as having one of the good charac- for the approved funds which makes this important reteristics listed in the IEEE-830 recommended practices search viable and efective. which is traceable as well as may help in fulfilling other requirements quality characteristics such as modifiable and verifiable. Thus, it is clear that the findings were References supporting H1.

Secondly, the findings from the analysis ascertained [1] IEEE Recommended Standard Practices for softthat a software development project is considered suc- ware Requirement Specification,8 30-std-1989 cessful when the project has fulfilled several criteria. The [2] I. Ozkaya, 2006, ’Representing Requirements Recriteria are listed as the following: 1) the outcome of the lationship’, F irst International Workshop on Reproject meets all the specified requirements; 2) the over- quirements Engineering Visualisation, IEEE Comall quality of the product is high; 3) The requirements- puter Society, pp. 3-3. related tasks (e.g. requirements specification, require- [3] L.Karlsson, A. Dahlstedt. B.Regnell. J.N.O. Dag A. ments management) are successfully completed in the Perrson 2007, ’Requirements Engineering chalproject; 4) The outcome of the project meets the business lenges in market driven software developmentgoal; and 5) The project is completed within scope. In- An interview study with practitioners’, Elsevier deed, all of the criteria listed are in fact quite similar to the Science Direct - Information and Software Techcriteria proposed by previous researchers (e.g. [ 10, 12 ]). nology, vol. 49, no. 2007, pp. 588-604. The finding asserted that as long as the outcome meets all [4] S. Diev, 2007, ’Structuring complex requirements’, the specified requirements and business goal, has good ACM SIGSOFT Software Engineering Notes Pages quality, completed within scope, and all the requirements- 1, vol. 32, no. 2, pp. 1-5. related activities are completed successfully, the project [5] G. Lucassen, F. Dalpiaz, 2016,’Improving agile rewill be considered successful although the project is not quirements:The quality user stories and tools’, in completed within time and budget. Thus, the findings J.M.E.M. van der Werf, et al. Requirements Eng are apparently shown how the success of requirements (2016) 21: 383. doi:10.1007/s00766-016-0250-x activities in which including maintaining requirements [6] P. Heck A. Zaidman, 2016,’ A Systematic Literaquality will impact success in which supporting H2. ture Review on Quality Criteria for Agile Require

References

vol. 24 ( 93 ), Springer, United State. [7] M.I. Kamata T. Tamai , 2007 ,'How Does Require-

15th IEEE International Requirements Engineer-

ing Conference, RE 2007, October 15 -19th, New

Delhi , India, pp. 69 - 78 [8] E. Knauss C.E. Boustani 2008 , 'Assessing the Qual-

ity of Software Requirements Specifications' , 16th

Conference , RE 2008 , 8 - 12 September, Barcelona,

Catalunya , Spain. IEEE Computer Society, pp. 341 -

342 [9] R. Berntsson Svensson , T. Olsson B. Regnell

( 2013 ). An investigation of how quality require-

and Software Technology , 55 ( 7 ), 1224 - 1236 . DOI:

10.1016/j.infsof. 2013 . 01 . 006 [10] C. Wohlin , A.V. Mayrhauser , M. Höst B. Regnell

Technology Vol . 42 , No. 14 , 2000 , pp. 983 - 992 [11] G. Thomas , W. Fernandez , 2008 'Success in IT

Journal of Project Management 26 ( 7 ), October,

2008 [12] N. Agarwal U . Rathod 2006 , 'Defining success

24 [13] M. Ibraigheeth S. A. Fadzli , 2019 , ' Core factor for

on Information visualisation ,Vol. 3 , No. 1 [14] V. Boola 2015 , Impact of project success factors in

analysis , Sage Journal , Vol ( 3 ), issue 2, pp: 109 - 125 . [15] M.H. D. Nasir S. Sahibuddin , 2011 ,'Critical suc-

study', Scientific Research and Essays Vol. 6 ( 10 ),

pp. 2174 - 2186 , 18 May, 2011 [16] K.E. Emam N.H. Madhavji , 1995 , 'Measuring the

IEEE , pp. 204 - 211 . [17] L. Radlinski , 2012 , ' Empirical Analysis of the im-

Bjorn , R. Daniela , D .(Eds), pp. 232 - 238 , Springer-

Verlag Berlin, Heidelberg. [18] A. Hussain , E.O. C. , Mkpojiogu F.M. Kamal , 2016 ,

management and marketing , Vol. 6 , special issue

(7). [19] J. Verner , K. Cox , S. Bleistein N . Cerpa 2005 , 'Re-

vol. 13 , no. 1 , pp. 225 - 238 . [20] I. Ozkaya O . Akin 2005 , 'Requirement-Driven de-

studies , vol. 27 , no. 2006 , pp. 381 - 398 . [21] A.M. Davis , 2005 , Just Enough Requirements Man-

Marketing , Dorset House Publishing, New York [22] S. Jungmayr , 2002 , 'Identfying Test-Critical Depen-

ence on Software Maintenance (ICSM'02) , IEEE

Computer Society . [23] A.G. Dahlstedt A . Persson 2005 , 'Requirements

95- 116 . [24] D.C. Howel , 2008 , The Analysis of Missing Data

Science Methodology , London, Sage. [25] J.F. Hair , W.C. Black , B.J.Babin R.E. Anderson 2010

NJ: Prentice-Hall. [26] R.B. Kline 1998 , Principles and Practice of Struc-

York. [27] W.W. Chin 1998 . Issues and opinion on structural

equation modeling . MIS Quarterly , 22 ( 1 ), 7 - 16 . [28] J.F. Hair , M. Sarstedt , C.M. Ringle J . A. Mena 2012 :

http://www.springerlink.com/content/t502155t60nv8005/ [29] , W.M.K. Trochim J.P. Donnelly 2008 , The

Atomic Dog a part of CENGAGE Learning , USA. [30] U. Sekaran and R. Bougie 2010 , Research Methods

Wiley and Sons. [31] D.W. Barclay , R. Thompson C.A. Higgins 1995 The

an Illustration . Tech Studies 2 ( 2 ): 285 - 309 . [32] D.R. Compeau , C.A. Higgins , S. Huf 1999 . So-

MIS Q 23 ( 2 ): 145 - 158 . [33] J. Kirk M.L. Miller 1986 , Reliability and Validity

in qualitative research , vol. 1 , Sage Publications ,

USA. [34] J. Nunnally L. Berstein 1994. Psychometric Theory .

New-York, NY: McGraw-Hill [35] C. Fornell D.F. Larcker , 1981 . Evaluating structural

measurement error . J Mark Res 18 ( 1 ): 39 - 50 . [36] P.M. Podsakof , D.W. Organ 1986 , ' Self-reports in

Journal of Management , 12 ( 4 ), 531 - 544 . [37] M. Wetzels , G.A. Schroder V.C. Oppen 2009 . Us-

tration. MIS Quarterly , 33 ( 1 ), 177 - 195 . [38] K , J.Preacher G.J. Leonardelli 2013 at

access on 13 February , 2013 [39] A.A. Ruhaya W. Bernard 2015 , 'The Interplay be-

(PLS)', Procedia of Computer Science Vol. 46 ,

No. 2015 , pp. 732 - 741