Table of Contents Author Guidelines Submit a Manuscript
Mobile Information Systems
Volume 2015 (2015), Article ID 649374, 9 pages
Research Article

Effects of Dexterity Level and Hand Anthropometric Dimensions on Smartphone Users’ Satisfaction

1Department of Industrial Engineering, Eskisehir Osmangazi University, Eskisehir, Turkey
2Statistics Department, Yıldız Technical University, Istanbul, Turkey

Received 8 June 2015; Accepted 12 October 2015

Academic Editor: Salil Kanhere

Copyright © 2015 N. Firat Ozkan and Fulya Gokalp-Yavuz. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


The usage of smartphones instead of simple mobile phones increases sharply in our era, especially among young people, because they do multiple tasks with single equipment. This study mainly focuses on smartphone satisfaction by combining hand measurements, smartphone users’ survey results, and hand dexterity levels of corresponding users acquired from Minnesota Manual Dexterity Test (MMDT). Structural Equation Modelling (SEM) is used as a statistical tool to discover the potential direct and indirect relations among user satisfaction, hand dimensions, and dexterity scores. Results indicates that thumb length, hand length, and dexterity level of the users have notable effects on users’ satisfaction with smartphones. Based on the results, a new approach that includes both gross motor skills and physical measurements is suggested to see hidden indirect relations with satisfaction.

1. Introduction

The increasing need for fast communication brings along the widespread use of the latest communication technologies. New forms of communication become mobile and they tend to coalesce into a single unit which is called smartphone. Numerous brands offer various smartphone models that have different technical features, physical designs, screen types, input devices, and so forth. People prefer different models of smartphones depending on their needs but it is quite hard to anticipate how they will be pleased with their smartphone. Smartphones have various features that are used in daily life such as standard phone calls, video phone calls, various instant messaging systems, advanced audio video recording technology, rapid Internet access, and various other features. Therefore, smartphone users have different expectations and purposes of using their devices. Their expectations and purposes of using affect their habits of the usage and the level of satisfaction. In the literature, there are numerous studies that have been conducted to measure user satisfaction and uncover information about the use of these devices.

Balakrishnan and Yeow [1] investigate the relationship between hand dimensions and short message service (sms) satisfaction. They measured hand breadth, thumb length, and thumb circumference measurements of the participants and applied a questionnaire to measure sms satisfaction. As a result of the study, it is suggested that manufacturing customized mobile phones for people who have larger thumbs can increase customer satisfaction.

Zulkefly and Baharudin [2] study the extent of mobile phone use amongst university students. They use several questionnaires to determine family and personal factors affecting purposes of using the mobile phone and its features using. Choi and Lee [3] focus on smartphone interface simplicity in their study. They conduct an online survey among smartphone users to evaluate their smartphone’s interface design in terms of simplicity. Park and Han [4] investigate the effects of touch key sizes and locations on one-thumb input on a mobile phone. They compare three different touch key sizes and twenty-five locations and use thumb length, thumb breadth, and hand length data of the participants. Lobo et al. [5] explain some guidelines which increase web usability of smartphones. Nitsche et al. [6] design an ergonomic user interface for a mobile search application by following a user centred design process which includes related questionnaires. They finalize the study with usability tests for their ergonomic user interface concept.

Physical design of smartphones is also a crucial point for user satisfaction. Since smartphones may require use of two hands and different fingers depending on the activity, smartphone sizes, screen, and keyboard sizes have a critical importance for the ease of usage. Many researchers conduct studies about physical design and mobile phone sizes and hand anthropometrics.

Jain and Pathmanathan [7] investigate keypad design satisfaction of mobile phone users using questionnaires, mobile phone dimensions, and 20 different hand measurements of the participants. Bradley et al. [8] conduct a survey on 362 people to examine user capabilities on mobile phone related tasks.

According to most of the related studies’ results, usability of devices and anthropometric features of the users are considered as critical points for design. These are supporting points to clarify the relationship between user expectations and device attributes. However, previous studies do not take into account the human capabilities that may affect overall user satisfaction on mobile devices. Essentially, it is very crucial to know approximate manual dexterity level of the target market for designing more appropriate devices. Although user capability is considered as a component of user satisfaction in some recent studies, there is still a gap in terms of considering motor skills of the users. This study fills this gap through using Minnesota Manual Dexterity Test (MMDT) as a part of the user satisfaction research.

Manual dexterity is a measurable characteristic and it is one of the indicators of human capabilities. Some tests are available to determine dexterity of one or two hands, but not both hands. Since smartphones may require use of both hands and several fingers, it is important to consider effect of user’s manual dexterity on their satisfaction of use. Since using the smartphone is not totally the same as using conventional mobile phones, it often requires use of both hands and fingers besides thumbs. On the other hand, smartphones offer much more features related with screen size and keyboard; because of that, their dimensions are bigger. At this point, manual dexterity level of users and choosing the most appropriate smartphone model must be emphasized in terms of user satisfaction. People should decide their smartphone model considering their aims of use, hand dimensions, and manual dexterity level. This study tries to emphasize the relationships between these three aspects and satisfaction level of smartphone users.

In this study, dexterity level is considered as an indirect effect on the satisfaction, besides hand anthropometric dimensions that directly (or naturally) affect dexterity level. It is possible to use both direct and indirect effects with the help of Structural Equation Modelling (SEM) as an extension of Path Analysis (PA) [9] to determine variables influencing the outcome (satisfaction). SEM encompasses PA and both models use the same underlying idea of model fitting and testing [10]. However, SEM allows us to work with latent variables that are weighted values of some observed satisfaction indicators. Also, it takes into account measurement errors and allows using mediator in the model [11]. The reason why SEM is preferred in this study is that it contains latent variables, contrary to PA, and it takes into account the measurement error, especially for independent variables, contrary to classical regression models.

The paper is structured as follows. Section 2 describes methods including user hand/finger dimensions, satisfaction, and dexterity test. The characteristics of the data and statistical analysis, basically PA and SEM, are defined in Section 3. Section 4 describes model settings and comparisons. Sections 5 and 6 include SEM result diagrams, conclusions, and future work, respectively.

2. Method

A multistage measurement process is designed to collect the data. The study is conducted with 36 participants. Firstly, each participant is asked to answer the questionnaire that includes questions about demographics, smartphone choices, habits, and satisfaction. Secondly, hand and finger dimensions of participants are measured. Finally, each participant performs the Minnesota Manual Dexterity Test.

2.1. Hand and Finger Dimensions

All hand and finger dimensions are measured for both right and left hands. However, not all of them are used in analysis part because there are high correlations between some of these measurements. Lafayette anthropometric tapes and small anthropometer are used for measuring the hand dimensions such as hand length, hand breadth, palm length, index finger length, index finger breadth, thumb length, and thumb breadth. Hand and finger dimensions (mm) are presented in Table 1.

Table 1: Hand and finger dimensions (mm).

2.2. Dexterity Test

Minnesota Manual Dexterity (MMD) Test includes several test methods. Two of them used in this study are the placing test performed by single hand and the turning test performed by two hands. Dexterity scores are determined based on task completion duration [12]. Both placing and turning tests are performed two times. Total trial times are used to determine percentile scale value that is provided by Examiner’s Manual of the Minnesota Manual Dexterity Test. To obtain a composite score, the average of two percentile values is calculated. Analyses are implemented separately with placing, turning, and composite test scores.

Before starting the MMD test, each participant is informed about the tasks of the test and they are allowed to get familiar with the test equipment. After they performed both placing and turning tests, completion time of each task is recorded. Dexterity scores are provided in Table 2.

Table 2: Minnesota Manual Dexterity Test results.

2.3. User Satisfaction Questionnaire

1–5 Likert scale is used to measure satisfaction level of participants with their smartphones. Additionally, demographic information, the reasons for choosing their smartphones, daily usage preferences, usage habits, and satisfaction questions (Table 3) are asked in the questionnaire. To make the data collection process easier and to have more reliable results, all the participants are asked for their voluntary consent.

Table 3: Satisfaction questions and rotated component matrix (rotation converged in 3 iterations).

3. Statistical Analysis

3.1. Data

The survey is conducted with 36 participants, the average age is 23 ranging between 19 and 34, and half of the participants are female. The main characteristics of the data are as follows: the average monthly income is 1007 (±796) TL (Turkish Lira), while the average family income is 4000 (±2153) TL. 83% of them are right-handed. The smartphones are used mostly for phone calls and instant messaging programs such as Whatsapp, Tango, with 39% and 25%, respectively. The brand and the price of the smartphone are two most popular answers, with 39% and 27%, respectively, to the question of “What is the most important feature for you while buying a smartphone?” The dimension of the smartphone is generally the second or the third option for the participants while buying a smartphone.

The objectives of smartphone usage and the satisfaction questions are asked with 5 Likert points with definitely dissatisfied to highly satisfied scale in the survey. Cronbach’s Alpha for the reliability of these items is 0.849. The average scores of satisfaction questions are mostly higher than 4 and this shows that participants are satisfied with their smartphones. However, the participants state that “it is possible to do some mistakes if they need to act faster,” which is one of the indicators of physical dissatisfaction. The average score for the related question is 2.67, that is, the lowest score among all satisfaction scores. The average scores obtained from the answers of questions M16, M17, M19, and M20 (descriptions are placed in Table 3) are lower than 4, as well.

Data collection part also includes hand/finger measurements and Minnesota Manual Dexterity Test results mentioned in Sections 2.1 and 2.2, respectively. The raw data and/or related information can be obtained from the author via e-mail.

3.2. Analysis

By using -test, it is concluded that there is no significant difference between right and left hand measurements with 5% confidence level. The analysis part proceeds with factor analysis. Participants answer several questions about their satisfaction in physical and general sense, and it is observed that the scores are split into two groups (Table 3). The first one is the physical satisfaction component, and the second one is the general satisfaction component. The final rotated component matrix of factor analysis is located at Table 3.

In the analysis part, it is aimed at keeping the number of variables limited, because of the sample size constraint. Additionally, since variables are highly correlated, two indicators for each factor are preferred. One is the physical satisfaction () composed of M16 and M17, and the other one is the phone size-general satisfaction () composed of M21 and M24. We implemented structural equational model (SEM) with two different types of indicators of satisfaction.

3.3. Path Analysis and Structural Equation Modelling

Path Analysis (PA), introduced by Wright [9], is a statistical tool to indicate direct and indirect relations between variables. The correlations among variables influencing the outcome are used to write structural equations in Wright’s analysis. Blau and Duncan [13] introduce PA into social scientific research. The extensive research on PA is developed by Blau and Duncan [13] in the book The American Occupational Structure. In their path models, they utilize occupational and educational outcomes in a sample of male adults and parents of them. In 1970s, PA gains its popularity among sociology, psychology, political science, economics, ecology, and other methods.

It is widely known that regression and correlation analysis are used to show relations between variables, but they are not quite enough to explain direct and indirect effects together. Pedhazur [14] emphasizes that “PA is intended not to discover causes but to shed light on the tenability of the causal models a researcher formulates based on knowledge and theoretical considerations” (p. 769).

Equations in Structural Equation Modelling (SEM) are known to be regression-like, because it is possible to use error terms which may not be independent of the other predictors like in regression models [15]. Maruyama [16] states that “Regression for prediction does not provide logic consistent with SEM approaches. The set of uses of regression in which the particular predictors and their regression weights are of interest, called regression for explanation, define why SEM techniques are so valuable” (p. 21).

Path diagram and path coefficients are two main tools for PA and also for SEM. Path diagram is the visual representation of the total effects of explanatory variables, and it consists of observed variables (rectangles) and latent variables (circles) connected by single-headed and double-headed arrows. It is mandatory to use double-headed arrows between exogenous variables which are assumed to have the variance explained by causes outside of the model. Conversely, endogenous variables’ variances are assumed to be explained by exogenous variables and other endogenous variables.

Figure 1 depicts an example of a PA model. Variables 1 and 2 are exogenous variables, while variables 3 and 4 are endogenous variables. “” and “” are residuals and they are not correlated. The arrows are drawn from the variables assumed as causes to variables assumed as effects [14, 17]. Path coefficient located on the diagram is shown by with “” subscript indicating the effect and “” indicating the cause. shows the correlation between exogenous variables. As it is seen in Figure 1, all variables are observed (not latent) in PA.

Figure 1: PA diagram example.

Figure 2 depicts an example of a SEM model. There are 4 observed variables denoted by that receive two paths going to them (the one from latent variable and the other one from their residual terms) and two latent variables denoted by and . () is error term and it is possible to use correlations among errors for SEM, although it is not mandatory. The same as in the path diagram, the curved two-way arrows show the correlation between variables.

Figure 2: SEM diagram example.

Model definition equations are main tools of SEM to see the relationship between observed and unobserved variables. Following Figure 2, the equations are written as follows:where , , denotes the factor loadings estimated based on the observed data and , , denotes residual terms. Detailed implementations for variances and covariance are found in related text books such as [10, 16, 18] among others. It is essential for SEM to use a computational program because of its burden mathematical complexity. We used AMOS [19] to implement our analysis. Programming languages such as AMOS use iteration for implementations, and it is crucial to check whether these minimization routines with iterations converge or not.

4. Model Setting and Comparisons

Before finalizing the model setting, a set of plausible models are tried. Maruyama’s quote [16] “attempting to impose a single path analytic solution to interpret makes no sense” (p. 18) is a common view in PA and SEM literatures. Also, using too many unknowns in a SEM model leads to not having an unequally solvable model. So, it is not preferable to add all variables existing in the study while building the model. There is also one thing to be kept in mind while determining the number of variables which is the sample size, which is recommended to be more than ten times the number of free parameters [20, 21].

In this study, it is not so easy to increase the sample size because of the expense of data collection process. Additionally, since there are high correlations between satisfaction indicators, two main indicators for satisfaction for both latent variables are used. shows the physical satisfaction indicator, while shows the phone size-general satisfaction for the rest of the study. Also, placing percentile scores (PPS) and turning percentile scores (TPS) are added separately as mediators to the models, since placing task is completed with single hand, while turning test is completed with both hands. This discrimination makes using left hand measurements for left-handed participants and right hand measurements for right-handed participants important, even though we did not find any statistical differences between the measurements of hand sizes.

According to satisfaction indicators and turning/placing percentile scores, four different models are implemented (Table 4). In Models 1 and 2, physical satisfaction () is taken as latent variable. The mediator of Model 1 is TPS, while the mediator of Model 2 is PPS. In Models 3 and 4, phone size-general satisfaction () is taken as latent variable. The mediator of Model 3 is TPS, while the mediator of Model 4 is PPS. Also, each model is built in two different versions. The first one (A) omits mediator and correlations of errors as in classical regression model and the second one (B) is named as default model that includes both mediator and correlations. Each model includes hand length and thumb length of participants as exogenous variables (Table 4).

Table 4: Model comparisons.

The first and the base model (Model 1-A, Table 4) does not include mediator or any covariance between independent variables and errors. The Chi-square test value of this model fit is 8.937 and probability value of the Chi-square test is 0.03 with 3 degrees of freedom (Table 4). An insignificant result at a 0.05 threshold is expected for a good model fit [22]. Therefore, in the base model, since value is smaller than the 0.05 level, the null hypothesis is rejected that model fits the data. Contrary to classical methodology in statistics analysis in which it is aimed at rejecting the null hypothesis (usually alternative hypothesis reflects the difference or change), SEM usually concerns not rejecting the null hypothesis, since it shows the model fits the data well [10].

Another absolute fit statistic is root mean square error of approximation (RMSEA) which is sensitive to the estimated number of parameters in the model [23]. Hooper et al. [24] report that it is taken as “one of the most informative fit indices” [25] (p. 54) because of its sensitivity property. Revisiting Model 1-A, RMSEA value (0.238) is not counted as good for expected well-fitted data since the model with a RMSEA greater than 0.1 would not be preferable [26]. Another fit index which takes into account both the measure of fit and model complexity is Akaike information criterion (AIC) [27] that is reported in the model comparisons table (AIC = 22.937 for the first model).

One of the main benefits of SEM is that it allows researchers to work with indirect effects of variables in addition to direct variables. These indirect effects show the effects between two variables that are mediated by one or more intervening variables (mediators) [10]. In the default version of Model 1 (B) with two degrees of freedom, the null hypothesis is not rejected that model fits the data well ( value: 0.779). Also, RMSEA value is acceptable.

It is noteworthy that Chi-square values and RMSEA get smaller values for all models with mediators. Additionally, SEM with mediators provides better results for general satisfaction indicator which is considered in Model 3 and Model 4. It is observed that, including a mediator to the models with is not as effected as . It is an inevitable result because physical satisfaction is directly affected more by both hand and thump lengths than the general satisfaction. This emphasizes the importance of mediator effect in indirect types of relations in SEM structures.

5. SEM Diagrams

The SEM diagrams of default version of four models (Model 1-B, Model 2-B, Model 3-B, and Model 4-B) with mediators and their comparisons are detailed in this section. The first theoretical model (Figure 3) depicts a mediated model, in which turning percentile scale modifies the effects of hand length and thumb length on which is a physical satisfaction indicator. It is pointed that hand and thumb lengths have direct effects, as well as having indirect effect on satisfaction. Hand length has a direct and positive (positive means that when hand lengths get bigger, physical satisfaction gets higher) effect on satisfaction with 0.97 standardized regression coefficient. See Figures 3, 4, 5, and 6.

Figure 3: SEM diagram of Model 1-B.
Figure 4: SEM diagram of Model 2-B.
Figure 5: SEM diagram of Model 3-B.
Figure 6: SEM diagram of Model 4-B.

In Figures 3, 4, 5, and 6, the correlations between exogenous variables are specified with double-headed arrows. Since the same exogenous variables are used for each model, the correlations () are all the same. The numbers on one-headed arrows indicate standardized regression coefficients. It is observed that the direct effect of hand and thumb lengths on physical satisfaction indicator is observed, in addition to the indirect effect which is the product of the path coefficients. For example, in Model 4-B (Figure 6), the direct effect of hand length on general satisfaction () is 0.61, while the indirect effect of hand length is .

6. Conclusion and Future Work

Smartphones are one of the most popular devices that are being used in daily life and requiring intensive human-machine interaction. Most of the consumers make their smartphone choices based on a few criteria. Usually, some technical features of smartphones and attractive appearance are being considered. This study attracts attention to the relationship between human attributes and satisfaction level of smartphones. Besides anthropometric dimensions, dexterity levels of participants are taken into account as a new approach and possible effects on smartphone satisfaction are inspected. Since motor skills of users are considered for the first time in this study, relevant findings provide a new viewpoint for smartphone satisfaction studies.

One of the main aims of this study is to investigate the satisfaction factors while selecting and using smartphones. For that purpose, first, hand anthropometric measurements of the participants were collected and then a multistage experimental study was conducted. Minnesota Manual Dexterity Test was utilized to make participants perform one-hand and two-hand dexterity tasks. A survey was conducted to assess satisfaction measurements. Based on survey results, it was possible to group the response under two main factors (physical satisfaction and general satisfaction). The factors in concern were hand anthropometric measurements and manual dexterity levels. Therefore, effects of these factors were investigated through detailed statistical analyses.

Another aim of this study is to assess the effects of placing percentile scale and turning percentile scale as mediators. Therefore, SEM is used as a statistical tool that is an extension of multiple regression models. After finalizing data collection, several models were implemented with two different types of indicators of satisfaction, which were determined with factor analysis, for different types of dexterity test results. Measured variables were selected as exogenous variables, while two variables were selected to generate latent variables for two different satisfaction indicators. Even though the direct effects of hand and thumb lengths were smaller on general satisfaction than physical satisfaction, it was possible to obtain the indirect effects of these lengths on general satisfaction with the help of SEM.

SEM indicates that standardized regression coefficients (Figures 3, 4, 5, and 6) between hand/thumb lengths and physical satisfaction () are higher than general satisfaction (), since includes direct physical satisfaction questions that are “I do not have difficulty in reaching keys with my thumbs” and “I push neither incorrect nor multiple keys with my thumbs.” Additionally, there is a negative relation between thumb length and satisfaction for each model; users with longer thumbs are less satisfied physically with their smartphones than the users with shorter thumbs. This result may bring an idea for smartphone practitioners and manufacturers to produce the new generation smartphones with varied key sizes. As it is known that there is a difference between male and female thumb sizes, key sizes may be specified for genders.

As a conclusion, it can be stated that dexterity level and hand anthropometrics of the users affect smartphone satisfaction either directly or latently. Analyses results indicate a critical importance of considering user attributes regarding dexterity and hand anthropometrics for both designers and consumers in the smartphone market.

Creating a more personal preference profile for users is an ongoing and natural extension of this study. Finding scientific evidence related to user-based factors for smartphone preferences would assist designers to invest more user friendly devices for their clients, especially in our era in which smartphones are tremendously popular and the sector of them are competitive indeed for designers and manufacturers. Future studies may consider different age groups. Further, other mobile devices used in daily life that require human-machine interaction may be studied. On the other hand, it is being planned developing a smartphone application to measure dexterity levels of potential users which will be applied prior to buying decision to help estimate prospective satisfaction level of them on the relevant smartphone.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.


  1. V. Balakrishnan and P. H. P. Yeow, “A study of the effect of thumb sizes on mobile phone texting satisfaction,” Journal of Usability Studies, vol. 3, no. 3, pp. 118–128, 2008. View at Google Scholar
  2. S. N. Zulkefly and R. Baharudin, “Mobile phone use amongst students in a university in Malaysia: its correlates and relationship to psychological health,” European Journal of Scientific Research, vol. 27, no. 2, pp. 206–218, 2009. View at Google Scholar · View at Scopus
  3. J. H. Choi and H.-J. Lee, “Facets of simplicity for the smartphone interface: a structural model,” International Journal of Human Computer Studies, vol. 70, no. 2, pp. 129–142, 2012. View at Publisher · View at Google Scholar · View at Scopus
  4. Y. S. Park and S. H. Han, “Touch key design for one-handed thumb interaction with a mobile phone: effects of touch key size and touch key location,” International Journal of Industrial Ergonomics, vol. 40, no. 1, pp. 68–76, 2010. View at Publisher · View at Google Scholar · View at Scopus
  5. D. Lobo, K. Kaskaloglu, C. Y. Kim, and S. Herbert, “Web usability guidelines for smartphones: a synergic approach,” International Journal of Information and Electronics Engineering, vol. 1, no. 1, pp. 33–37, 2011. View at Publisher · View at Google Scholar
  6. M. Nitsche, A. Nürnberger, and K. Bade, “An ergonomic user interface supporting information search and organization on a mobile device,” in Proceedings of the Personal Information Management in a Socially Networked World, Seattle, Wash, USA, February 2012.
  7. S. Jain and G. Pathmanathan, “Importance of anthropometry for designing user-friendly devices: mobile phones,” Journal of Ergonomics, vol. 2, no. 4, 2012. View at Publisher · View at Google Scholar
  8. M. Bradley, S. Waller, J. Goodman-Deane et al., “A population perspective on mobile phone related tasks,” in Designing Inclusive Systems, pp. 55–64, Springer, London, UK, 2012. View at Google Scholar
  9. S. Wright, “Correlation and causation,” Journal of Agricultural Research, vol. 20, no. 7, pp. 557–585, 1921. View at Google Scholar
  10. T. Raykov and G. A. Marcoulides, A First Course in Structural Equation Modeling, Lawrance Erlbaum Associates, 2000.
  11. R. M. Baron and D. A. Kenny, “The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations,” Journal of Personality and Social Psychology, vol. 51, no. 6, pp. 1173–1182, 1986. View at Publisher · View at Google Scholar · View at Scopus
  12. T. I. M. Hilgenkamp, R. Van Wijck, and H. M. Evenhuis, “Physical fitness in older people with ID—concept and measuring instruments: a review,” Research in Developmental Disabilities, vol. 31, no. 5, pp. 1027–1038, 2010. View at Publisher · View at Google Scholar · View at Scopus
  13. P. Blau and O. Duncan, The American Occupational Structure, Wiley, 1967.
  14. E. L. Pedhazur, Multiple Regression in Behavioral Research, Thomson Learning, 1997.
  15. R. P. McDonald and M.-H. R. Ho, “Principles and practice in reporting structural equation analyses,” Psychological Methods, vol. 7, no. 1, pp. 64–82, 2002. View at Publisher · View at Google Scholar · View at Scopus
  16. G. M. Maruyama, Basics of Structural Equation Modeling, Sage Publications, 1997.
  17. S. Wright, “The method of path coefficients,” The Annals of Mathematical Statistics, vol. 5, no. 3, pp. 161–215, 1934. View at Publisher · View at Google Scholar
  18. R. B. Kline, Principles and Practice of Structural Equation Modeling, Guilford Press, New York, NY, USA, 2005.
  19. J. L. Arbuckle, Amos User's Guide, Smallwaters, Chicago, Ill, USA, 1995.
  20. P. M. Bentler, EQS Structural Equations Program Manual, Multivariate Software, Encino, Calif, USA, 2006.
  21. L.-T. Hu, P. M. Bentler, and Y. Kano, “Can test statistics in covariance structure analysis be trusted?” Psychological Bulletin, vol. 112, no. 2, pp. 351–362, 1992. View at Publisher · View at Google Scholar · View at Scopus
  22. P. Barrett, “Structural equation modelling: adjudging model fit,” Personality and Individual Differences, vol. 42, no. 5, pp. 815–824, 2007. View at Publisher · View at Google Scholar · View at Scopus
  23. J. H. Steiger, “Structural model evaluation and modification: an interval estimation approach,” Multivariate Behavioral Research, vol. 25, no. 2, pp. 173–180, 1990. View at Publisher · View at Google Scholar
  24. D. Hooper, J. Coughlan, and M. R. Mullen, “Structural equation modelling: guidelines for determining model fit,” The Electronic Journal of Business Research Methods, vol. 6, no. 1, pp. 53–60, 2008. View at Google Scholar · View at Scopus
  25. A. Diamantopoulos and J. A. Siguaw, Introducing LISREL, Sage, London, UK, 2000.
  26. M. W. Browne and R. Cudeck, “Alternative ways of assessing model fit,” Sociological Methods & Research, vol. 21, no. 2, pp. 230–258, 1992. View at Publisher · View at Google Scholar
  27. H. Akaike, “Factor analysis and AIC,” Psychometrika, vol. 52, no. 3, pp. 317–332, 1987. View at Publisher · View at Google Scholar · View at Scopus