You are on page 1of 10

A Research about Measurement Invariance of Attitude

Participating in Field Hockey Sport


Dr. Mao-Chun Chiu, Department of Leisure, Recreation & Health Business Management,
Tajen University, Taiwan

ABSTRACT
The research goal is to test the measurement invariance of attitude participating in field hockey
sport. By utilizing cluster sampling, three hundred and seventy-nine valid samples were selected and
analyzed for descriptive analysis and structural equation model by statistical software SPSS12.0 and
AMOS 16.0. Results revealed: First, there was no difference between the expected covariance equation
and sample covariance. Second, there were no significant differences among measurement model,
structural model, and covariance model after cross-validity analysis. At last, concrete suggestions are
provided to government and other hockey sport organizations for further improvement.
Keywords: Field hockey, attitude toward sport participation
INTRODUCTION
Doing exercise is effective for health improvement. With the economic growth in Taiwan, the
awareness of leisure improves in the public, and the rapid increase of their participation in leisure and
sports. Field hockey is an ancient sport with long history; the ball is small with high speed and is played
in a spacious field with players high skills (Tsai, 1996). Field hockey is suitable for eastern countries
because of its limitlessness of body type and gender. Being included in the competition events in Asian
Games, field hockey has potential for improvement in Taiwan. Ice hockey and field hockey are two kinds
of hockey sport, and the latter is suitable to be promoted in Taiwan for its warm weather. The field,
number of players, and game tactics are similar to those of soccer game; therefore, field hockey has
become a popular leisure sport of the public.
Myers (1993) addressed that ones attitude is formed by his consciousness, emotion, and behavioral
tendency. Attitude toward exercise is, regardless of results, ones or most peoples gain to their health
physically, satisfaction, happiness, and spirituality rely on adequate time and space with their independent
willingness. Attitude refers to an individuals constancy and coherence tendency towards human affairs
and his/her surroundings. This tendency could be predicted by ones external behavior, but the
connotation not only means external behavior, attitude generally includes consciousness, emotion, and
behavior (Chang, 1989). Hence, for the measurement of attitude toward exercise, it could be individually
analyzed ones attitude by feelings of doing exercise, belief, and tendency toward exercise; whether their
attitude are positive or negative, liking or disliking, participating it or not. By realizing ones attitude
researchers could predict his level of preference to the consciousness, favor and behavior of one matter
(Wang, Chen, 2006).
Ajzen and Fishbeins Reasoned Action Theory which addressed in 1980 indicated that attitude plays
an important role in constructing behavioral intention, and intention is a norm in predicting the
occurrence of actual behavior. Ajen (1988) considered attitude as a reflection tendency from some
particular people whether they continuously like or dislike what they learned, Promoting sport

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

61

participating behavior has been a focus worldwide for decades, relevant theory indicated that attitude is a
crucial factor affecting an individuals behavior and activity (Noland & Feldman, 1985).
The previous studies mainly focused on the impacts of participation hindrance to leisure behavior
(Huang, Chen, 2005; Alexandris, Tsorbatoudis & Grouios, 2002; Alexandris & Carroll, 1997; Alexandris
& Carroll, 1999). This research attempts to utilize a strict testing process to construct a model of attitude
toward participating field hockey sport. Through a series of testing analysis on factor structure, a most
parsimonious factor model is expected. Reliability, convergent validity, and discriminant validity of this
factor model will be examined; also, the measurement invariance of attitude participating in field
hockey sport will be tested by cross validity.
RESEARCH METHOD
Research Structure
Based on relevant references, this research selected quantification research methods and tested the
results by Structural Equation Modeling. Statistical Model is shown in diagram 1.
1
e8

1
e7

1
e6

1
e5

1
e4

1
e3

1
e2

1
e1

1
e23

1
e22

1
e21

1
e20

I1
B1
1

B2
B3
Benefits

B5

I5

Involvement

I6

B6

I7
1

I8

B8

I9
I10

A1

I11

Ability and
skill

A2
A3
A4

I3
I4

B4

B7

I2

Achievement and
Satisfaction

S1
S2
S3
S4
S5

1
e9

1
e10

1
e11

1
e12

1
e13

1
e14

1
e15

1
e16

1
e17

1
e18

1
e19

1
e24

1
e25

1
e26

1
e27

1
e28

Diagram 1: Statistical Model


Research Hypothesis
Structural Equation Model (SEM) use co-variance model to estimate the mutual-support and
correlative relationship among several Multi-Regression Equations. (Lee, Ko, Wu, Yu, 2004; Jreskog &
Srbom ,1992) Chen (2007) emphasized that structural equation model adopted the covariant equation
testing variable relationship among the variables, the diversity between the covariance equation and
sample covariance equation of the received model ought to be smaller, the smaller the better. Chin
(1998) indicated that the goodness of fit of SEM is not allowed to be inferior. It refers to the obvious
disparity between the model and the sample and the wrong model design, and will result in the following
incorrect research outcomes. Hence, this researchs first step is to advance the hypothesis to the goodness

62

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

of fit of this model, S-()=0, S is the sample covariance equation, and () is the model expected
covariance equation.
H1: There is no difference between the model expected covariance equation of the model and the sample
covariance equation.
H2: The model of attitude toward participating in filed hockey sport possesses measurement invariance.
Research Tool
(1) Questionnaire Design
Two sections in this research questionnaire are: the scale about the model of attitude toward
participating in field hockey sport and personal information.
1. Questionnaire Model about attitude participating in field hockey sport
This questionnaire was revised from Yang and Kus research and their scale of university students
attitude toward leisure sport in 2004. This questionnaire includes twenty-eight questions in re sport
benefit, sport involvement, ability and skill, achievement and satisfaction respectively.
2. Descriptive Statistics
This questionnaire was revised from Yang and Kus research and their scale of university students
attitude toward leisure sport in 2004. A descriptive analysis includes gender, group, and annual income
respectively.
(2) Likert Scaling
Bollen (1989) indicated that seven-point Likert scale in SEM practically reveals the best performance,
and therefore was selected for the questionnaire scoring. Four questionnaire items are benefits from sport,
investment on sport, ability and skills, achievement and psychological satisfaction; and categories of this
scale are rated from Strongly agree (seven points) to Strongly disagree (one point).
Sampling
The duration for data collection was from August, 5th to 13th, 2014 with random sampling of
subjects as participants in National Presidential In-door Hockey Cup in 2014. Those participants were
questioned an hour before the competitions which began in Changhua Stadium, and 400 questionnaires
were collected.
Sample Estimation and Statistical Power
After settling the SEM model, the amount of samples needs to be determined before sample
collection. Based on the RMSEA estimation method provided by MacCallum, Browne and Sugawara in
1966, and by utilizing the minimum sample size computed by R language and the degree of freedom 344
in this survey for estimation, the minimum sample size is 88.67. The effective sample size in this survey
is 379 and it fits the suggested value mentioned above. Also, the statistical power is 1, which is bigger
than the suggested value 0.80. (Maxwell, Kelley, & Rausch, 2008) Results revealed that the statistical
power is ideal.
RESULTS
Descriptive Statistics of Sample Characters
Among the effective samples in this research, 52.2 percent of subjects are males and 47.8 percent
are females; which refers to 198 and 181 respectively. In the case of group, 105 subjects were studying
in elementary school, 126 in the junior high school, 87 in senior high school, and 61 in the adult group.

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

63

As for the item income, in 323 peoples monthly salary is under twenty thousand dollars, 37 people are
in the group between 20,001 to 30,000; only 19 people earn more than thirty thousand dollars.
Table 1: Summary of Descriptive Statistics about Sample Characters
Information Item
Category Standard
Sample size
Male
198
Gender
Female
181
Elementary School
105
Junior High School
126
Group
Senior High School
87
Adult
61
Under 20,000
323
Income
20,001-30,000 dollars
37
Above 30,001
19

Percentage
52.2
47.8
27.7
33.2
23.0
16.1
85.2
9.8
5.0

Measurement and Structural Model Analysis


(1) Confirmation of Convergent Validity
Confirmatory Factor Analysis (CFA) is a crucial step of SEM analysis. This survey, was amended based
on the two-stage model from Kline in 2005, downsized the variables of CFA measurement model. First, test the
measurement model before implementing the structural model evaluation. If the goodness of fit of the
measured model is acceptable, complete SEM model evaluation will be conducted in the second stage.
This research is aimed to conduct CFA analysis on the four aspects in this model as the benefits
from sport, investment on sport, ability and skills, achievement and psychological satisfaction. Factor
loading of all aspects are among 0.74 to 0.94; all reach the significant standard. Composite reliability is
between the number of 0.95 to 0.98; and the average variance extracted is between the number of 0.78 to
0.83 (shown in Table 2), which fits the suggested value addressed by Hair, Anderson, Tatham and Black
in 1998, and by Fornell and Larcker in 1981.

Aspect

Benefits

Involvement

64

Table 2: Reliability Analysis of Potential Aspects


StandarNon-StanardC.R.
Norm
S.E.
dized Loading dized Loading
(t-value)
B1
0.90
1.00
B2
0.88
0.97
0.04 26.40
B3
0.91
1.01
0.03 28.75
B4
0.86
0.96
0.04 25.01
B5
0.94
1.02
0.03 31.18
B6
0.94
1.01
0.03 31.58
B7
0.94
1.01
0.03 31.15
B8
0.90
0.96
0.03 28.15
I1
0.85
1.00
I2
0.89
1.06
0.04 24.30
I3
0.92
1.07
0.04 25.67
I4
0.94
1.07
0.04 27.03
I5
0.89
1.06
0.04 24.11
I6
0.90
1.04
0.04 24.52
I7
0.92
1.05
0.04 25.52
I8
0.85
1.09
0.05 21.98
I9
0.92
1.03
0.04 25.64

P
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***

SMC C.R. AVE


0.81 0.97 0.83
0.77
0.82
0.74
0.88
0.88
0.88
0.81
0.73 0.98 0.80
0.80
0.84
0.89
0.79
0.81
0.84
0.72
0.84

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

Ability and skill

Achievement
and Satisfaction

I10
I11
A1
A2
A3
A4
S1
S2
S3
S4
S5

0.85
0.91
0.92
0.92
0.91
0.88
0.74
0.92
0.91
0.91
0.92

0.95
1.13
1.00
0.98
0.99
0.91
1.00
1.08
1.04
1.12
1.02

0.04
0.05

22.04
25.02

0.03
0.03
0.03

31.03
30.06
27.12

0.06
0.06
0.06
0.05

18.99
18.73
18.75
18.88

*** 0.72
*** 0.82
0.84 0.95 0.82
*** 0.84
*** 0.82
*** 0.77
0.54 0.95 0.78
*** 0.84
*** 0.82
*** 0.82
*** 0.84

(2)Confirmation of Discriminant Validity


The aim of discriminant validity analysis is to confirm whether the difference occurs to two
different aspects. This survey chose confidence interval method (Torkzadeh, Koufteros, Pflughoeft, 2003)
to confirm the confidence interval of correlation coefficient between the two aspects. If the interval
excludes 1 (which reaches the complete correlation), discriminant validity exists between aspects. To
establish a confidence interval of correlation coefficient in SEM under the circumstance of confidence
level of 95%, this survey utilized Bootstrap estimation. If the confidence interval excludes 1, null
hypotheses are refused, and the four aspects possess discriminant validity. Hancock and Nevitt (1999)
suggested that the number of Bootstrapping should be more than two hundred and fifty times when
estimating the path coefficient. This survey, with a 95% confident level, repeatedly used sampling over
one thousand times to retrieve the confidence interval of standardized correlation coefficient.
AMOS bootstrap provides estimation for two confidence intervals, which are Bias-corrected
Percentile Method and Percentile Method. Results of these two estimations are shown in Table 3all
standardized confidence interval of correlation coefficient exclude 1, which means discriminant validity
exists between one aspect and another.
Table 3: Bootstrap Correlation Coefficient 95% Confidence Interval
Bias-corrected Percentile method
Parameter
Estimate
Lower Upper Lower
Upper
Benefits
<-->
Ability and skill
.93
.90
.95
.90
.95
Benefits
<-->
Involvement
.92
.88
.94
.87
.94
Benefits
<--> Achievement and Satisfaction
.93
.90
.95
.89
.95
Involvement <-->
Ability and skill
.95
.91
.97
.91
.97
Ability and skill <--> Achievement and Satisfaction
.97
.95
.99
.95
.99
Involvement <--> Achievement and Satisfaction
.95
.91
.97
.91
.97
(3) Structural Model Analysis
If the sample size of SEM is over 200, it normally causes bigger chi-squared value.
(2=(n-1)Fmin). Fmin is the minimum value of difference between sample equation and expectation
equation. The bigger the sample size it is, the higher the chi-squared value it becomes. P value therefore
becomes easily be refused (Chang, 2011). For this reason, Bollen and Stine addressed Bootstrape for
amendment in 1992.
The chi value of Bollen-stine p correction is 527.94, and the original ML chi value is 2399.92. Because
the chi-value became smaller, all goodness of fit requires computation again. Results are listed in Table 4.
Structural model analysis includes model fitness and explanatory power of entire research model. This

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

65

research took Bagozzi and Yi (1988), Bentler (1995), and Hair (1988)s opinions, using seven indexes for
evaluation for goodness of fit of over-all model, including chi-value test (2), 2 and ratio of degree of freedom,
goodness of fit index (GFI), adjusted goodness of fit index (AGFI), root mean square error of approximation
(RMSEA), comparative fit index (CFI), and normed fit index (NFI). Results are listed in Table 6.
Bagozzi and Yi (1988) emphasized that, taking 2 and ratio of degree of freedom to test the
goodness of fit, the degree is ideal to be smaller. The ratio of this research is 3(1.53); Hair, etc. (1988)
indicated that GFI and AGFI are ideal to close to 1, no absolute standard to judge the goodness of fit.
Baumgartner and Homburg (1996) recommended that the values of GFI and AGFI in the research need to
be higher than 0.90. GFI and AGFI in this research model are 0.97 and 0.96 respectively. Browne and
Cudeck (1993) explained a good model with reasonable fir if its RMSEA is lower than 0.08. The RMSEA
of this research model is 0.04, the allowable standard is 0.90. CFI in this research model is 0.99. NFI
value as least needs to be higher than 0.90, while the NFI is 0.97 in this research model. Over all, the fit
indices are within the standard value, which showed the research results are acceptable. Also, the data of
the samples in this research can be used to explain practical observation data.
Table 4: Fitness Analysis of Research Model
Allowable Range
Research Model
the smaller the better
527.94
378
<3
1.53
>0.9
0.97
>0.9
0.96
<.08
0.04
>0.9
0.99
>0.9
0.97

Fit Indices
2(Chi-square)
DF
2/DF
GFI
AGFI
RMSEA
CFI
NFI

Judgment of model fit

Fit
Fit
Fit
Fit
Fit
Fit
I1

.80

B1

e8

I2

.77

B2

e7

.90
.88
.91
.86

.83

B3

e6

.73

B4

e5

.88

.85
.89
.92
.94
.90

.94

Benefits

B5

e4

.88
e3

B6

.94
.84
.93

.91

B7
.83

.93

B8

e1

A1

.92
.92

.85

A2

e22

.80

.90

A3

e21

.95
.93

.85
e23

.90

.80

A4

I5
I6

.94

.88
e2

e20

I4
.89

Involvement

.92

I3

.93

I8

.86
.95

I7

I9

.90

I10
I11

Ability and
skill
.97

Achievement and
Satisfaction

S1

.71
.91

S2
.92

.93 .92

S3
S4
S5

.72
e9

.79
e10

.84
e11

.88
e12

.81
e13

.80
e14

.86
e15

.70
e16

.86
e17

.73
e18

.81
e19

.51
e24

.83
e25

.85
e26

.84
e27

.87
e28

Diagram 2: Model of Attitude Participating in Field Hockey Sport

66

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

(4) Cross Validity


This research is operated under the circumstance that the assumption of the research model is
accurate, and two groups are compared after random sampling. (Cliff, 1983; Cudeck & Broene, 1983;
Hairs, etc., 1983) To deeply test the stability of this model, this research progressively restrict three
coefficients, including path coefficient of the measured model, path coefficient of structure and structural
covariance. If no obvious differences are found in this model, it owns a certain level of stability.
1. Set equal path coefficient to two groups, twenty four factor loadings in total are designed equally in
structural model (DF=24) and Chi-square measure (CMIN) increased 12.83. Besides, the test result
p=.97 and failed to reach the significant level, which means setting twenty four factor loading
equally is acceptable.
2. Besides remaining the restriction of the measurement model, plus setting ten structural path
coefficient
(DF=34-24=10),
Chi-square
measure
(CMIN)
increased
18.06
(CMIN=30.89-12.83=18.06). The test result p=.62 and failed to reach the significant level; in other
words, these ten structural path coefficient are acceptable, and are the same.
3. With the same restriction to the structural efficient model, and extra twenty eight set structural
variances and covariances (DF=62-34=28). Chi-square measure increased 109.74
(CMIN=140.63-30.89=109.74).The test result p= 0.00 and has reached the significant level 0.05;
which means making these twenty eight variances and covariances equal is unacceptable. These
twenty eight variances and covariances are unequal.
Chung and Rensvold (2002) brought up the practical significance of test based on CFI norm.
Results showed CFI0.01 after the simulation, which means that no differences between the two nested
structural model. Nevertheless, Little (1997) addressed that TLI0.05 is the standard of having no
difference among nested structural model. The invariance comparison of CFI and TLI both fit the
suggested value mentioned as well as the requirement of congruent groups. This model is stable and fits
the standard of cross validity.
Table 5: Comparison of group interval invariance
Model

df

Measurement weights
Structural covariances
Measurement residuals

12.83
30.89
140.63

24
34
62

df
10
28

CFI

18.06
109.74

.97
.62
.00

.87
.87
.87

CFI

TLI

.00
.00

-.01
-.01
-.01

CONCLUSION
Most subjects in this survey are males, currently in junior high school, and monthly income lower
than 20,000. This research includes three stages and questionnaires are analyzed by confirmatory factor
analysis. Research results revealed that this questionnaire includes benefit, involvement, achievement and
satisfaction, ability and skill, four factors with twenty eight questions in total. For the application of
structural equation model, Bollen (1989) considered that Likert seven-point scale is ideal to reduce the
over-skewed results, and confirmatory research is suitable because the variance level is greater. Therefore,
the questionnaire design selected was the seven-point scale.
After analyzed by composite reliability, average variance extracted, this questionnaire possesses
good convergent validity. With bootstrap confident interval, two estimated methodsBias-corrected
Percentile Method and Percentile Method were used to estimate disterminant validity. Results showed

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

67

that every potential variable in this questionnaire is distinguishable. Seven scales in this research were
evaluated for overall model fit, including 2 test, ratio of 2 and degree of freedom, goodness of fit (GFI),
adjusted goodness of fit index (AGFI), root mean square error of approximation (RMSEA), and
comparative fit index (CFI), parsimony-adjusted comparative fit index (PCFI), results showed good
model fit. Therefore, the first hypothesis that there is no difference between the model expected
covariance equation of the model and the sample covariance equation is valid.
This research randomly conducted cross-validity confirmation of two groups, both CFI and TLI
as the invariance comparison are smaller than the standard suggested by the scholars, and therefore fit the
requirement of equal group interval. This model has stability and fit the cross-validity. As the result, the
second hypothesis that the model of attitude toward participating in field hockey sport possesses
measurement invariance is valid.
SUGGESTIONS
Academic contribution of this research
In the past thirty years, structural equation model has become a widely used statistical technology.
Researchers applied SEM to construction models and attempted to know the relationship among variables.
By operating the model to test the relationship of hypotheses, and further collect data for evaluation.
According to Schreiber (2008), McDonald and Ho (2002), Boomsma (2000), and Hoyle and Panter (1995)
s suggestions, well-designed SEM research articles must display sample size, statistical power, the
version of statistical software used (AMOS 16.0), goodness of fit, Chi-square measurement, multi fit
indices (GFI, AGFI, CFI, NFI, RMSEA), parameter estimation of measurement and structure including
standardized and non-standardized estimated value and the reports of significance, SMC and explainable
variance, final statistical model diagram, and cross validity. Based on the recommendation, this research
provided a more complete statistical analysis report for other researchers in the future.
Statistical power evaluation is very crucial work, because sample size plays the key role while
testing the model. (Bollen, 1989) The degree of freedom in this research shows the minimum number of
the sample size needs to be higher than 344; the valid sample size is 396 and has reached its standard.
Under the request of statistical power, the general statistical power is ideal to be 0.8 (Maxwell, Kelley &
Rausch, 2008), and the statistical power in this research is 1.0 and is ideal for statistical analysis.
Suggestions for Field Hockey Sport Participation
Field hockey sport helps relieve players study pressure and tension, this research recommends
schools and relevant organizations to form field hockey clubs for promotion and pressure relief. Although
the number of hockey players in Taiwan is lower than it in other countries; the researchers expect that by
the active promotion of sports fair council can field hockey sport become popular in Taiwan and further
produce positive impact on players skill and improvement. In the trend of globalization, the demarcation
lines among nations are blurred and bring more frequent skill interactions in countries. Inviting coaches
from other countries for player training in Taiwan to improve competition level is also highly
recommended.

68

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

REFERENCES
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall.
Alexandris, K. & Carroll, B. (1997). Demographic differences in the perception of constraints on recreational sport participation:
Results from a study in Greece. Leisure Studies, 16, 107-125.
Alexandris, K. & Carroll, B. (1999). Constraints on recreational sport participation in adults in Greece: Implications for providing
and managing sport services. Journal of Sport Management, 13, 317-332.
Alexandris, K., Tsorbatzoudis, C. & Grouios, G. (2002). Perceived constraints on recreational sport participation: Investigating their
relationship with intrinsic motivation, extrinsic motivation and motivation. Journal of Leisure Research, 34(3), 233-252.
Bagozzi, R. P., & Yi, Y. (1988). On the evaluation for structural equation models. Journal of the Academy of Marketing Science, 16,
74-94.
Baumgartner, H., & Homburg, C. (1996). Applications of structural equation modeling in marketing and consumer research: A
review. International Journal of Research in Marketing, 13(2), 139-161.
Bentler, P. M. (1995). EQS: Structural equations program manual. Encino, CA: Multivariate Software .
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Bollen, K. A., & Stine, R. A.(1992). Bootstrapping goodness-of fit measures in structural equation models. Sociological methods
and Research,21,205-229.
Boomsma, A. (2000). Reporting analyses of covariance structures. Structural Equation Modeling, 7, 461483.
Brown, M. W., & Cudeck, R. (1993). Alternatives ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing
structural equation models. Newbury Park, CA: Sage
Chang, W. H.(2001). Structural Equation Modeling. Kaohsiung: Tristar.
Chen, S.Y. (2007). Structural Equation Modeling Amos Operation. Taipei: Psychological Publishing Co.
Cheung, G. W., & Rensvold, R. B. (2002).Evaluating goodness-of-fit indexes for testing measurement invariance. Structural
Equation Modeling, 9(2), 233-255.
Chin, W. W. (1998). Issues and opinions on structural equation modeling, MIS Quarterly. 22 (1), 7-16.
Chung, C.H. (1989). Chengs Psychology Dictionary. Taipei: Tung Hua Publisher.
Cliff, N. (1983). Some cautions of causal modeling methods. Multivariate Behavioral Research, 18, 115-126.
Cudeck, R., & Browne, M. W. (1983). Cross validation of covariance structures. Multivariate Behavioral Research, 18,147-167.
Fornell, C., & Lacker, D. F.(1981). Evaluating structural equation models with unobservable variables and measurement error,
Journal of Marketing Research. 18, 39-50.
Hair, J. F. Jr., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis (5th ed.). Upper saddle River, NJ:
Prentice Hall.
Hancock, G. R., & Nevitt, J. (1999).Bootstrapping and the identification of exogenous latent variables withi n structural equation
models. Structural Equation Modeling, 6, 394-399.
Hoyle, R. H., & Panter, A. T. (1995). Writing about structural Equation models. In R. H. Hoyle (Ed.). Structural equation modeling:
Concepts, issues, and applications, Thousand Oaks, CA: Sage, 158-176.
Huang, Y. L., & Chen, H. M.(2005). The Relationship between Attitude and Recreational Sport Behavior .Journal of outdoor
recreation,18(3), 81-102.
Jreskog, K. G., & Srbom, D. (1992). LISREL: A guide to the program and applications (3rd ed.). Chicago: Scientific Software
International Research, 32, 5376.
Lee, N. H., Koo, T. Y., & Wu, G. S., (2004). Construction of the Behavioral Tendency Model of Tourist in Kinmen, Journal of
Management, 21(1), 131-151.
Little, T. D. (1997). Mean and covariance structures (MACS) analyses of cross-cultural data: Practical and theoretical issues.
Multivariate Behavioral Research, 32, 53-76.

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

69

MacCallum, R.C., Browne, M. W., & Sugawara, H. M.(1996). Power analysis and determination of sample size for covarianc e
structure modeling. Psychological Methods, 1(2),130-149.
Maxwell, S. E., Kelley, K., & Rausch, J. R. (2008). Sample size planning for statistical power and accuracy in parameter esti mation.
Annual Review of Psychology, 59,537-563.
McDonald, R. P., & Ho, M. H. R. (2002). Principles and practice in reporting structural equation analyses. Psychological Methods, 7, 6482.
Myers, D. G. (1993). Social Psychology(4th ed.). NY: McGraw-Hill.
Noland, M. P. & Feldman, R. H. L.(1985). An empirical investigation of leisure exercise behavior in adult women. Health Education,
16, 29-34.
Schreiber, J. B. (2008). Core reporting practices in structural equation modeling. Administrative Pharmacy, 4, 83-97.
Tsai, Y. M. (1996) . Physical application in Field Hockey. Taipei Physical Education College Press, 5, 139-147.
Torkzadeh, G., Koufteros, X., & Pflughoeft, K. (2003), Confirmatory analysis of computer self-efficacy. Structural Equation
Modeling, 10(2), 263-275.
Wang, J. M., & Chen, C. C.(2006).The effects of coach efficacy and leadership on athletes' sport motivation and sport attitude.
Bulletin of Sports and Exercise Psychology, 9, 1-18.
Yang, L. M., & Ku, Y. C.(2004). The development and application of exercise attitude scale for university student. Physical
Educational Journal, 37,149-161.

Appendix 1: Covariance Matrix


B1
2.203
1.826
1.758
1.694
1.737
1.862
1.755
1.722
1.392
1.421
1.693
1.713
1.632
1.582
1.668
1.474
1.684
1.675
1.644
1.639
1.595
1.590
1.569
1.171
1.570
1.793
1.648
1.666

70

B2
1.826
2.163
1.770
1.702
1.732
1.720
1.680
1.695
1.377
1.449
1.562
1.543
1.548
1.453
1.585
1.382
1.569
1.481
1.528
1.594
1.553
1.541
1.538
1.311
1.545
1.722
1.606
1.557

B3
1.758
1.770
2.162
1.713
1.861
1.763
1.808
1.714
1.475
1.535
1.562
1.637
1.686
1.540
1.662
1.450
1.608
1.491
1.605
1.715
1.784
1.656
1.659
1.437
1.696
1.662
1.803
1.560

B4
1.694
1.702
1.713
2.213
1.773
1.677
1.688
1.710
1.305
1.407
1.443
1.432
1.529
1.429
1.570
1.371
1.526
1.469
1.469
1.551
1.513
1.549
1.548
1.202
1.491
1.627
1.567
1.532

B5
1.737
1.732
1.861
1.773
2.107
1.824
1.869
1.729
1.498
1.607
1.637
1.643
1.785
1.608
1.728
1.575
1.692
1.549
1.667
1.723
1.719
1.672
1.596
1.482
1.633
1.669
1.764
1.642

B6
1.862
1.720
1.763
1.677
1.824
2.046
1.854
1.704
1.375
1.468
1.596
1.658
1.649
1.580
1.658
1.447
1.604
1.549
1.618
1.661
1.660
1.629
1.567
1.198
1.533
1.734
1.650
1.624

B7
1.755
1.680
1.808
1.688
1.869
1.854
2.050
1.663
1.534
1.620
1.651
1.628
1.774
1.656
1.685
1.545
1.624
1.465
1.761
1.784
1.713
1.716
1.578
1.438
1.660
1.658
1.788
1.623

B8
1.722
1.695
1.714
1.710
1.729
1.704
1.663
1.987
1.444
1.503
1.593
1.578
1.620
1.546
1.700
1.480
1.632
1.598
1.565
1.667
1.665
1.675
1.666
1.276
1.581
1.704
1.673
1.584

I1
1.392
1.377
1.475
1.305
1.498
1.375
1.534
1.444
2.125
1.857
1.740
1.628
1.571
1.667
1.574
1.717
1.446
1.307
1.824
1.595
1.499
1.535
1.514
1.538
1.500
1.464
1.659
1.533

I2
1.421
1.449
1.535
1.407
1.607
1.468
1.620
1.503
1.857
2.187
1.801
1.723
1.716
1.654
1.700
1.874
1.658
1.450
1.927
1.818
1.655
1.589
1.551
1.624
1.652
1.540
1.767
1.654

I3
1.693
1.562
1.562
1.443
1.637
1.596
1.651
1.593
1.740
1.801
2.080
1.827
1.665
1.646
1.723
1.747
1.654
1.577
1.883
1.761
1.668
1.608
1.549
1.470
1.604
1.659
1.776
1.637

I4
1.713
1.543
1.637
1.432
1.643
1.658
1.628
1.578
1.628
1.723
1.827
1.998
1.757
1.775
1.724
1.783
1.682
1.578
1.849
1.702
1.745
1.624
1.540
1.317
1.531
1.596
1.746
1.656

I5
1.632
1.548
1.686
1.529
1.785
1.649
1.774
1.620
1.571
1.716
1.665
1.757
2.194
1.700
1.820
1.790
1.738
1.677
1.795
1.838
1.793
1.757
1.630
1.551
1.696
1.604
1.845
1.652

I6
1.582
1.453
1.540
1.429
1.608
1.580
1.656
1.546
1.667
1.654
1.646
1.775
1.700
2.078
1.684
1.791
1.654
1.490
1.874
1.574
1.652
1.586
1.540
1.353
1.544
1.588
1.711
1.593

I7
1.668
1.585
1.662
1.570
1.728
1.658
1.685
1.700
1.574
1.700
1.723
1.724
1.820
1.684
2.044
1.789
1.749
1.658
1.720
1.754
1.818
1.691
1.736
1.513
1.712
1.733
1.863
1.636

I8
1.474
1.382
1.450
1.371
1.575
1.447
1.545
1.480
1.717
1.874
1.747
1.783
1.790
1.791
1.789
2.564
1.744
1.467
1.984
1.668
1.681
1.599
1.546
1.465
1.596
1.569
1.752
1.545

I9
1.684
1.569
1.608
1.526
1.692
1.604
1.624
1.632
1.446
1.658
1.654
1.682
1.738
1.654
1.749
1.744
1.945
1.676
1.789
1.688
1.688
1.611
1.576
1.456
1.638
1.656
1.717
1.638

I10
1.675
1.481
1.491
1.469
1.549
1.549
1.465
1.598
1.307
1.450
1.577
1.578
1.677
1.490
1.658
1.467
1.676
1.948
1.579
1.516
1.513
1.496
1.483
1.235
1.496
1.597
1.528
1.510

I11
1.644
1.528
1.605
1.469
1.667
1.618
1.761
1.565
1.824
1.927
1.883
1.849
1.795
1.874
1.720
1.984
1.789
1.579
2.376
1.845
1.756
1.705
1.583
1.463
1.689
1.614
1.791
1.678

A1
1.639
1.594
1.715
1.551
1.723
1.661
1.784
1.667
1.595
1.818
1.761
1.702
1.838
1.574
1.754
1.668
1.688
1.516
1.845
2.213
1.888
1.868
1.649
1.583
1.768
1.666
1.834
1.753

A2
1.595
1.553
1.784
1.513
1.719
1.660
1.713
1.665
1.499
1.655
1.668
1.745
1.793
1.652
1.818
1.681
1.688
1.513
1.756
1.888
2.132
1.795
1.677
1.528
1.710
1.633
1.928
1.647

A3
1.590
1.541
1.656
1.549
1.672
1.629
1.716
1.675
1.535
1.589
1.608
1.624
1.757
1.586
1.691
1.599
1.611
1.496
1.705
1.868
1.795
2.231
1.770
1.515
1.713
1.637
1.825
1.647

A4
1.569
1.538
1.659
1.548
1.596
1.567
1.578
1.666
1.514
1.551
1.549
1.540
1.630
1.540
1.736
1.546
1.576
1.483
1.583
1.649
1.677
1.770
2.010
1.447
1.661
1.696
1.840
1.593

S1
1.171
1.311
1.437
1.202
1.482
1.198
1.438
1.276
1.538
1.624
1.470
1.317
1.551
1.353
1.513
1.465
1.456
1.235
1.463
1.583
1.528
1.515
1.447
2.821
1.878
1.390
1.802
1.510

S2
1.570
1.545
1.696
1.491
1.633
1.533
1.660
1.581
1.500
1.652
1.604
1.531
1.696
1.544
1.712
1.596
1.638
1.496
1.689
1.768
1.710
1.713
1.661
1.878
2.104
1.732
1.873
1.662

S3
1.793
1.722
1.662
1.627
1.669
1.734
1.658
1.704
1.464
1.540
1.659
1.596
1.604
1.588
1.733
1.569
1.656
1.597
1.614
1.666
1.633
1.637
1.696
1.390
1.732
2.001
1.788
1.709

S4
1.648
1.606
1.803
1.567
1.764
1.650
1.788
1.673
1.659
1.767
1.776
1.746
1.845
1.711
1.863
1.752
1.717
1.528
1.791
1.834
1.928
1.825
1.840
1.802
1.873
1.788
2.336
1.780

S5
1.666
1.557
1.560
1.532
1.642
1.624
1.623
1.584
1.533
1.654
1.637
1.656
1.652
1.593
1.636
1.545
1.638
1.510
1.678
1.753
1.647
1.647
1.593
1.510
1.662
1.709
1.780
1.912

The Journal of Human Resource and Adult Learning, Vol. 10, Num. 2, December, 2014 issue

You might also like