5 confirmatory factor analysis -...
TRANSCRIPT
132
5 Confirmatory Factor Analysis
133
5.1 INTRODUCTION
The hypothesized model of the study consists of 11 factors which were
originally measured using a questionnaire of 57 statements. Previous analysis of the
collected data resulted into the elimination of 4 items and hence, 53 items were taken
further for confirmatory factor analysis. This chapter presents the confirmatory
analysis carried out before the hypothesized model is tested for its fitness.
The objective of this chapter is to develop the measurement model of the study
variables. First section of the chapter presents an overview of CFA while the second
section describes the procedure adopted for conducting analysis along with the criteria
used for improving the measurement model. The third section of the chapter describes
the CFA for individual factors. Finally, the fourth section presents the CFA for overall
model.
5.2 CFA - OVERVIEW
Confirmatory factor analysis (CFA) is a statistical technique used to verify the
factor structure of a set of observed variables. In a CFA the researcher has a strong
idea about the number of factors, the relations among the factors, and the relationship
between the factors and measured variables. In exploratory factor analysis, factors
underlying a set of scale items are not known while extracting the factors. In
confirmatory factor analysis, the loading of scale items on factors are already known
through the theory and they are tested for such loadings to fit an observed set of data.
CFA is commonly used for establishing the validity of a single factor model, test the
significance of a specific factor loading, test whether a set of factors are correlated or
uncorrelated and assess the convergent and discriminant validity of a set of measures
(DeCoster, 1998).
134
5.3 ANALYSIS PROCEDURE
The measurement model is evaluated in two steps.
Step 1 CFA is conducted for each identified factor of the measurement model.
Step 2 CFA is conducted for all factors at the same time for getting the final
measurement model
In step 2, individual scale items were loaded on their appropriate factors and
all factors were correlated with each other.
5.4 EVALUATION CRITERIA FOR MEASUREMENT MODEL
The measurement model in step one and two are evaluated using the following
model fit indices as proposed by Byrne (2001), Hair, Anderson, Tatham, & Black
(1998) and Kumar (2010):
(i). 2 test
(ii). 2/df ratio
(iii). Comparative Fit Index (CFI)
(iv). Goodness-of-Fit Index (GFI)
(v). Root Mean Square Error of Approximation (RMSEA)
For improving the model, following test statistics are used (Kumar, 2010):
(i). Standardized Regression Weights (SRW)
(ii). Standardized Residual Covariances (SRC)
(iii). Modification Indices (MI)
135
2 2/df ratio are highly sensitive to the sample size In case of large
sample, these two statistics were not applicable to assess the fitness of the study
model (Garson, 2010; Ullman, 1996; Schumacker & Lomax, 2004). So, CFI, GFI and
RMSEA are utilized to assess the model fitness as fit indices.
The acceptance rules of the above indices and statistics are shown in the
Tables 5.1 and 5.2 respectively.
Table 5.1 Model Fit Indices
Fit Index Values Model Fitness
CFI
< 0.80 Unacceptable
0.80 to 0.90 Acceptable
> 0.90 Good
GFI
< 0.80 Unacceptable
0.80 to 0.90 Acceptable
> 0.90 Good
RMSEA
< 0.05, Good
< 0.08 Acceptable
< 0.10 Mediocre
poor
Table 5.2 Model Evaluation Test Statistics
Statistic Value Criteria Model Improvement
SRW Acceptable Not Required
< 0.4 Unacceptable Required
SRC - 2.58 to 2.58 Acceptable Not Required
Otherwise Unacceptable Required
MI < 10 Acceptable Not Required
Unacceptable Required
136
Table 5.1 is used for individual factors and overall model fitness and Table 5.2
is used to improve the model. Model improvement is made by eliminating or
correlating the measurement item which has low SRW, high SRC, and high MI.
5.5 MEASUREMENT MODEL TESTING
5.5.1 CFA for Each Factor
CFA was conducted for the following eleven individual factors of the
measurement model:
i) pleasure, ii) arousal, iii) merchandise quality, iv) service quality, v) perceived price, vi) ambient lighting,
vii) background music, viii) ambient scent, ix) in-store signage and graphics,
x) store satisfaction and xi) store loyalty.
5.5.1.1 Pleasure Factor CFA
The pleasure factor consists of five measurement items (PL2, PL3, PL4, PL5,
PL6) for CFA. The results of the CFA for pleasure model are shown in Table 5.3.
Table 5.3 CFA Results for Pleasure Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 5 384.218(5) 76.844 0.785*** 0.796*** 0.353***
Item PL5 is eliminated
in modification
Refined Model Results
4 126.083(2) 63.041 0.907* 0.912* 0.319***
Final Refined Model Results
4 10.842(1) 10.842 0.993* 0.991* 0.096**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
CFI, GFI and RMSEA of the initial model are unacceptable. So, model
improvement is required. For improving the model, item PL5 is eliminated based on
the statistics SRC and MI as reported in Tables 5.5 and 5.6. The SRC between PL5
137
and PL4 is high (9.283) and MI between PL4 and PL5 are also high (136.948 and
139.921) and unacceptable. Hence, the item PL5 is eliminated for improvement.
Table 5.4 Standardized Regression Weights for Pleasure Scale Items
SRW
PL6 <--- Pleasure 0.860
PL5 <--- Pleasure 0.549
PL4 <--- Pleasure 0.536
PL3 <--- Pleasure 0.796
PL2 <--- Pleasure 0.886
Table 5.5 Standardized Residual Covariance among Pleasure Scale Items
PL2 PL3 PL4 PL5 PL6
PL2 0.000 PL3 -0.901 0.000 PL4 1.624 -1.664 0.000 PL5 0.821 -0.754 9.283 0.000 PL6 0.079 1.636 -3.041 -2.538 0.000
Table 5.6 Modification Indices among Pleasure Scale Items
MI
PL2 <--- PL4 23.491
PL3 <--- PL4 10.817
PL3 <--- PL6 12.985
PL4 <--- PL5 136.948
PL4 <--- PL6 16.407
PL5 <--- PL4 139.921
PL5 <--- PL6 11.765
PL6 <--- PL3 20.988
PL6 <--- PL4 59.888
PL6 <--- PL5 42.032
138
CFA was conducted after eliminating the item PL5. The model fit results are
shown in Table 5.3. The RMSEA of refined model is unacceptable. So, further
improvement of the model is required. For improving the model PL2 and PL4 are
correlated based on SRC and MI statistics (Table 5.8 and 5.9). The SRC is high
(3.746) between PL2 and PL4 and the MI values are also high (75.230 and 20.840)
between PL2 and PL4. Hence, PL2 and PL4 were correlated.
Table 5.7 Standardized Regression Weights of Pleasure Scale Items
SRW
PL6 <--- Pleasure 0.906
PL4 <--- Pleasure 0.455
PL3 <--- Pleasure 0.820
PL2 <--- Pleasure 0.847
Table 5.8 Standardized Residual Covariance among Pleasure Scale Items
PL2 PL3 PL4 PL6
PL2 0.000 PL3 -0.682 0.000 PL4 3.746 -0.443 0.000 PL5 -0.036 0.437 -1.980 0.000
Table 5.9 Modification Indices among Pleasure Scale Items
MI
PL2 <--- PL4 75.230
PL4 <--- PL2 20.846
PL6 <--- PL4 45.660
After correlating PL2 and PL4, the CFA was conducted again. As shown in
Table 5.3, the model fit indices of final model i.e., CFI, GFI and RMSEA, indicate
139
that the refined pleasure factor is acceptable for structural equation model. The path
diagram for refined pleasure factor is shown in Figure 5.1.
Figure 5.1 Path Diagram for Refined Pleasure Factor
5.5.1.2 Arousal Factor CFA
The arousal factor consists of five measurement items (AR1, AR2, AR3, AR4,
AR6) for CFA. The results of the CFA for arousal model are shown in Table 5.10.
Table 5.10 CFA Results of Arousal Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 5 237.964(5) 47.593 0.772*** 0.871** 0.276*** Item AR6 is
eliminated in
modification Final Refined Model Results 4 22.601(2) 11.301 0.972* 0.983* 0.083**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
Except for GFI, CFI and RMSEA indicate unacceptable values for model
fitness. So, model improvement is required. For improving the model, item AR6 is
eliminated based on the statistic SRW as reported in Table 5.11. The value of SRW
for AR6 is 0.388 which is lower than the cut off value of 0.4. Hence, item AR6 is
eliminated for improvement.
140
Table 5.11 Standardized Regression Weights for Arousal Scale Items
SRW
AR6 <--- Arousal 0.388
AR4 <--- Arousal 0.517
AR3 <--- Arousal 0.819
AR2 <--- Arousal 0.632
AR1 <--- Arousal 0.792
Table 5.12 Standardized Residual Covariance among Arousal Scale Items
AR1 AR2 AR3 AR4 AR6
AR1 .000 AR2 -1.499 .000 AR3 .837 .328 .000 AR4 .171 .228 -1.538 .000 AR6 -1.752 5.258 -2.769 7.506 .000
Table 5.13 Modification Indices among Arousal Scale Items
MI
AR1 <--- AR2 10.516
AR1 <--- AR6 12.583
AR2 <--- AR6 55.015
AR3 <--- AR4 12.862
AR3 <--- AR6 38.938
AR4 <--- AR6 85.195
AR6 <--- AR2 35.572
AR6 <--- AR3 10.246
AR6 <--- AR4 71.128
After deleting item AR6, CFA was conducted again. The values of CFI, GFI
and RMSEA, as shown in Table 5.10, for the refined model indicate that the improved
141
arousal factor is acceptable. The path diagram for refined arousal factor is shown in
Figure 5.2.
Figure 5.2 Path Diagram for Refined Arousal Factor
5.5.1.3 Merchandise Quality Factor CFA
The merchandise quality factor consists of six measurement items (MQ1,
MQ2, MQ3, MQ4, MQ5, MQ6) for CFA. The results of the CFA for merchandise
quality model are shown in Table 5.14.
Table 5.14 CFA Results for Merchandise Quality Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 6 96.395(9) 10.711 0.933* 0.949* 0.126*** Item MQ3 is
eliminated in
modification Final Refined Model Results 5 24.917(4) 6.229 0.983* 0.985* 0.093**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
The CFI and GFI values indicate that the merchandise quality factor has a
good model fit. But RMSEA value indicates unacceptability of the model fitness.
Hence, model improvement is carried out by eliminating item MQ3 and correlating
items MQ1 and MQ2. The item MQ3 is eliminated because of low SRW of 0.384
(Table 5.15) which is below the cut off value of 0.4. The items MQ1 and MQ2 are
correlated because the SRC and MI values between MQ1 and MQ2 are well beyond
the cut off values (Tables 5.16 and 5.17).
142
Table 5.15 Standardized Regression Weights for Merchandise Quality Scale Items
SRW
MQ6 <--- Merchandise Quality 0.795
MQ5 <--- Merchandise Quality 0.783
MQ4 <--- Merchandise Quality 0.819
MQ3 <--- Merchandise Quality 0.384
MQ2 <--- Merchandise Quality 0.585
MQ1 <--- Merchandise Quality 0.554
Table 5.16 Standardized Residual Covariance among Merchandise Quality Scale Items
MQ1 MQ2 MQ3 MQ4 MQ5 MQ6
MQ1 0.000 MQ2 4.310 0.000 MQ3 -0.085 1.730 0.000 MQ4 -0.284 0.195 0.970 0.000 MQ5 -1.052 -2.006 -1.176 0.344 0.000 MQ6 -0.378 -0.205 -0.683 -0.476 0.949 0.000
Table 5.17 Modification Indices among Merchandise Quality Scale Items
MI
MQ1 <--- MQ2 31.486
MQ2 <--- MQ1 33.538
MQ5 <--- MQ2 15.986
After deleting the item MQ3 and correlating items MQ1 and MQ2, CFA was
conducted again. The results of the improved model are shown in Table 5.14. The
values of CFI, GFI and RMSEA of the refined merchandise quality factor are
acceptable for the structural equation model. The path diagram for refined
merchandise quality factor is shown in Figure 5.3.
143
Figure 5.3 Path Diagram for Refined Merchandise Quality Factor
5.5.1.4 Service Quality Factor CFA
The service quality factor consists of five measurement items (SQ1, SQ2,
SQ3, SQ4, SQ5) for CFA. The results of the CFA for service quality model are shown
in Table 5.18.
Table 5.18 CFA Results for Service Quality Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 5 96.250(5) 19.250 0.903* 0.937* 0.173*** No item is
eliminated in
modification Final Refined Model Results 5 11.643(3) 3.881 0.991* 0.992* 0.069**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
The model fitness results for CFI and GFI show good fit of the service quality
model while RMSEA indicate unacceptable model fitness. Hence, model
improvement is required. Based on the values of SRCs and MIs, model improvement
is done by correlating items SQ2 with SQ3 and SQ1 with SQ4. As shown in Table
5.20, the SRC between SQ2 and SQ3 is 2.769 and SRC between SQ1 and SQ4 is
2.622 which are above the cut off value of 2.58. The values of MI between SQ1 and
SQ4 are 14.212 and 20.607 and between SQ2 and SQ3 are 16.708 and 28.486 which
are also high (Table 5.21). So, SQ1 is correlated with SQ4 and SQ2 is correlated with
SQ3.
144
Table 5.19 Standardized Regression Weights for Service Quality Scale Items
SRW
SQ5 <--- Service Quality 0.680
SQ4 <--- Service Quality 0.714
SQ3 <--- Service Quality 0.754
SQ2 <--- Service Quality 0.612
SQ1 <--- Service Quality 0.600
Table 5.20 Standardized Residual Covariance among Service Quality Scale Items
SQ1 SQ2 SQ3 SQ4 SQ5
SQ1 0.000 SQ2 -1.591 0.000 SQ3 -1.876 2.769 0.000 SQ4 2.622 -2.003 -0.368 0.000 SQ5 0.799 -0.397 -0.309 0.134 0.000
Table 5.21 Modification Indices among Service Quality Scale Items
MI
SQ1 <--- SQ4 14.212
SQ2 <--- SQ3 16.708
SQ3 <--- SQ1 12.992
SQ3 <--- SQ2 28.486
SQ4 <--- SQ1 20.607
SQ4 <--- SQ2 12.095
After correlations are made, CFA was conducted. The improved fit indices for
refined service quality factor are shown Table 5.18. CFI, GFI and RMSEA values for
the refined model show acceptable fit for the improved service quality model. The
path diagram for refined service quality factor is shown in Figure 5.4.
145
Figure 5.4 Path Diagram for Refined Service Quality Factor
5.5.1.5 Perceived Price Factor CFA
The perceived price factor consists of four measurement items (PP1, PP2, PP3,
PP4) for CFA. The results of the CFA for perceived price model are shown in Table
5.22.
Table 5.22 CFA Results for Perceived Price Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 4 11.810(2) 5.905 0.991* 0.990* 0.090** No model
improvement required Final Refined
Model Results 4 -- -- -- -- --
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
CFI, GFI and RMSEA values show good fit for perceived price factor. Hence,
model improvement is not required for perceived price factor. The path diagram for
the perceived price is shown in Figure 5.5.
Figure 5.5 Path Diagram for Perceived Price Factor
146
5.5.1.6 Ambient Lighting Factor CFA
The ambient lighting factor consists of seven measurement items (AL1, AL2,
AL3, AL4, AL5, AL6, AL7) for CFA. The results of the CFA for ambient lighting
model are shown in Table 5.23.
Table 5.23 CFA Results for Ambient Lighting Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 7 198.162(14) 14.154 0.906* 0.917* 0.147***
Item AL1 is eliminated
in modification
Refined Model Results 6 80.654(9) 8.962 0.954* 0.957* 0.114***
Final Refined Model Results 6 45.250(8) 5.656 0.976* 0.976* 0.087**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
The model fitness results for CFI and GFI show good fit of the ambient
lighting model while RMSEA value indicate unacceptable model fitness. Hence,
model improvement is required. For improving the model, item AL1 is deleted based
on SRC and MI values. As can be seen in Table 5.25, SRC values associated with
AL1 are high (2.866 with AL2 and -3.155 with AL3). The MI values associated with
AL1 are also high (18.812 and 23.784 with AL2, 22.02 and 22.069 with AL3 and
12.562 with AL4). So, AL1 is eliminated from the model.
Table 5.24 Standardized Regression Weights for Ambient Lighting Scale Items
SRW
AL7 <--- Ambient Lighting 0.841
AL6 <--- Ambient Lighting 0.738
AL5 <--- Ambient Lighting 0.607
AL4 <--- Ambient Lighting 0.743
AL3 <--- Ambient Lighting 0.650
AL2 <--- Ambient Lighting 0.722
AL1 <--- Ambient Lighting 0.649
147
Table 5.25 Standardized Residual Covariance among Ambient Lighting Scale Items
AL1 AL2 AL3 AL4 AL5 AL6 AL7
AL1 0.000 AL2 2.866 0.000 AL3 -3.155 0.487 0.000 AL4 -1.984 -1.035 2.189 0.000 AL5 0.942 -2.727 0.091 1.689 0.000 AL6 0.622 -1.035 -0.035 -0.057 1.509 0.000 AL7 0.196 0.762 -0.318 -0.050 -0.778 -0.131 0.000
Table 5.26 Modification Indices among Ambient Lighting Scale Items
MI
AL1 <--- AL2 18.812
AL1 <--- AL3 22.020
AL2 <--- AL1 23.784
AL2 <--- AL5 21.062
AL3 <--- AL1 22.069
AL3 <--- AL4 11.129
AL4 <--- AL1 12.562
AL4 <--- AL3 15.303
AL5 <--- AL2 15.013
After eliminating AL1, CFA was conducted. The improved model fit indices
is shown in Table 5.23. Since RMSEA values indicate unacceptable model fitness,
further improvement is required. For further improving the model, items AL2 and
AL7 are correlated based on MI values in Table 5.29.
148
Table 5.27 Standardized Regression Weights of Ambient Lighting Scale Items
SRW
AL7 <--- Ambient Lighting 0.831
AL6 <--- Ambient Lighting 0.732
AL5 <--- Ambient Lighting 0.606
AL4 <--- Ambient Lighting 0.770
AL3 <--- Ambient Lighting 0.680
AL2 <--- Ambient Lighting 0.691
Table 5.28 Standardized Residual Covariance among Ambient Lighting Scale Items
AL2 AL3 AL4 AL5 AL6 AL7
AL2 0.000 AL3 0.454 0.000 AL4 -0.942 1.256 0.000 AL5 -2.301 -0.315 1.343 0.000 AL6 -0.455 -0.445 -0.376 1.620 0.000 AL7 1.475 -0.724 -0.351 -0.618 0.136 0.000
Table 5.29 Modification Indices among Ambient Lighting Scale Items
MI
AL2 <--- AL5 13.419
AL5 <--- AL2 10.576
AL7 <--- AL2 12.851
CFA was conducted again after the items AL7 and AL2 are correlated. Table
5.23 shows the refined model fit indices. CFI, GFI and RMSEA values show that the
refined ambient lighting model has acceptable fit. The path diagram for the refined
ambient lighting factor is shown in Figure 5.6.
149
Figure 5.6 Path Diagram for Refined Ambient Lighting Factor
5.5.1.7 Background Music Factor CFA
The background music factor consists of four measurement items (BM2, BM3,
BM4, BM5) for CFA. The results of the CFA for background music model are shown
in Table 5.30.
Table 5.30 CFA Results for Background Music Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 4 13.093(2) 6.546 0.995* 0.989* 0.095** No model
improvement required Final Refined
Model Results 4 -- -- -- -- --
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
CFI, GFI and RMSEA values show good fit for background music factor.
Hence, model improvement is not required for background music factor. The path
diagram for the background music is shown in Figure 5.7.
Figure 5.7 Path Diagram for Background Music Factor
150
5.5.1.8 Ambient Scent Factor CFA
The ambient scent factor consists of four measurement items (AS1, AS2, AS3,
AS4) for CFA. The results of the CFA for ambient scent model are shown in Table
5.31.
Table 5.31 CFA Results for Ambient Scent Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 4 9.584(2) 4.792 0.992* 0.992* 0.079** No model
improvement required Final Refined
Model Results 4 -- -- -- -- --
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
CFI, GFI and RMSEA values show good fit for ambient scent factor. Hence,
model improvement is not required for ambient lighting factor. The path diagram for
the ambient scent is shown in Figure 5.8.
Figure 5.8 Path Diagram for Ambient Scent Factor
5.5.1.9 In-store Signage and Graphics Factor CFA
The in-store signage and graphics factor consists of four measurement items
(SG2, SG3, SG4, SG5) for CFA. The results of the CFA for in-store signage and
graphics model are shown in Table 5.32.
151
Table 5.32 CFA Results for In-store Signage and Graphics
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 4 6.224(2) 3.112 0.995* 0.995* 0.059** No model
improvement required Final Refined
Model Results 4 -- -- -- -- --
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
CFI, GFI and RMSEA values show good fit for in-store signage and graphics
factor. Hence, model improvement is not done for in-store signage and graphics
factor. The path diagram for the in-store signage and graphics is shown in Figure 5.9.
Figure 5.9 Path Diagram for In-store Signage and Graphics
5.5.1.10 Store Satisfaction Factor CFA
The store satisfaction factor consists of four measurement items (SS1, SS2,
SS3, SS4) for CFA. The results of the CFA for store satisfaction model are shown in
Table 5.33.
Table 5.33 CFA Results for Perceived Price Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 4 7.885(2) 3.942 0.995* 0.994* 0.069** No model
improvement required Final Refined
Model Results 4 -- -- -- -- --
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
152
CFI, GFI and RMSEA values show good fit for store satisfaction factor.
Hence, model improvement is not required for store satisfaction factor. The path
diagram for the store satisfaction model is shown in Figure 5.10.
Figure 5.10 Path Diagram for Store Satisfaction Factor
5.5.1.11 Store Loyalty Factor CFA
The store loyalty factor consists of five measurement items (SL1, SL2, SL3,
SL4, SL5) for CFA. The results of the CFA for store loyalty model are shown in
Table 5.34.
Table 5.34 CFA Results for Store Loyalty Factor
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results 5 199.204(5) 39.841 0.906* 0.884** 0.252***
Item SL4 is eliminated
in modification
Refined Model Results 4 36.808(2) 18.404 0.976* 0.973* 0.169***
Final Refined Model Results 4 2.591(1) 2.591 0.999* 0.998* 0.051*
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
The model fitness results for CFI and GFI show good and acceptable fits of
the store loyalty model while RMSEA indicate unacceptable model fitness. Hence,
model improvement is required. For improving the model, item SL4 is eliminated.
Item SL4 is eliminated based on the values of SRC and MI as shown in Tables 5.36
and 5.37. The SRC value between SL2 and SL4 (2.883) is high. The MI values
153
associated with SL4 (10.635 with SL1, 46.258 and 32.817 with SL2) are also high.
So, item SL4 is eliminated for improvement.
Table 5.35 Standardized Regression Weights for Store Loyalty Scale Items
SRW
SL5 <--- Store Loyalty 0.786
SL4 <--- Store Loyalty 0.769
SL3 <--- Store Loyalty 0.830
SL2 <--- Store Loyalty 0.828
SL1 <--- Store Loyalty 0.855
Table 5.36 Standardized Residual Covariance among Store Loyalty Scale Items
SL1 SL2 SL3 SL4 SL5
SL1 0.000 SL2 -0.008 0.000 SL3 1.019 -1.473 0.000 SL4 -1.227 2.883 -1.026 0.000 SL5 -0.223 -0.661 1.231 -0.451 0.000
Table 5.37 Modification Indices among Store Loyalty Scale Items
MI
SL1 <--- SL4 10.635
SL2 <--- SL3 12.651
SL2 <--- SL4 46.258
SL3 <--- SL2 12.828
SL4 <--- SL2 32.817
CFA was conducted after eliminating SL4. The improved model fit statistics
are shown in Table 5.34. Except RMSEA, CFI and GFI values indicate the store
loyalty model is fit. Hence, further improvement is required. For improving the
model, items SL2 and SL3 are correlated based on MI values as shown in Table 5.40.
154
The modification index between SL2 and SL3 is 17.554 which is higher than cut off
value of 10. So, SL2 and SL3 are correlated
Table 5.38 Standardized Regression Weights for Store Loyalty Scale Items
SRW
SL5 <--- Store Loyalty 0.789
SL3 <--- Store Loyalty 0.860
SL2 <--- Store Loyalty 0.769
SL1 <--- Store Loyalty 0.881
Table 5.39 Standardized Residual Covariance among Store Loyalty Scale Items
SL1 SL2 SL3 SL5
SL1 0.000 SL2 0.640 0.000 SL3 0.074 -0.932 0.000 SL5 -0.688 0.261 0.684 0.000
Table 5.40 Modification Indices among Store Loyalty Scale Items
MI
SL2 <--- SL1 10.482
SL3 <--- SL2 17.554
SL5 <--- SL1 13.558
SL5 <--- SL3 10.605
CFA was conducted again after correlating the items SL2 and SL3. The
improved model indices of CFI, GFI and RMSEA depicted in the Table 5.34 indicate
that the final refined store loyalty model is acceptable for structural equation model.
The path diagram for refined store loyalty factor is shown in Figure 5.11.
155
Figure 5.11 Path Diagram for the Refined Store Loyalty Factor
A summary of the CFA results done on each of the factors is shown in Table
5.41 below. The fit statistics reported in the table are of the refined models of each
factor.
Table 5.41 Fit Statistics of CFA for Each Factor - Refined Models
Factor No. of Scale Items
2 (df) 2/df CFI GFI RMSEA
Refined Model (Total No. of Scale Items = 48)
Pleasure 4 10.842(1) 10.842 0.993 0.991 0.096
Arousal 4 22.601(2) 11.301 0.972 0.983 0.083 Merchandise Quality 5 24.917(4) 6.229 0.983 0.985 0.093
Service Quality 5 11.643(3) 3.881 0.991 0.992 0.069
Perceived Price 4 11.810(2) 5.905 0.991 0.990 0.090
Ambient Lighting 6 45.250(8) 5.656 0.976 0.976 0.087
Background Music 4 13.093(2) 6.546 0.995 0.989 0.095
Ambient Scent 4 9.584(2) 4.792 0.992 0.992 0.079 In-store Signage and Graphics 4 6.224(2) 3.112 0.995 0.995 0.059
Store Satisfaction 4 7.885(2) 3.942 0.995 0.994 0.069
Store Loyalty 4 2.591(1) 2.591 0.999 0.998 0.051
5.5.2 CFA for Overall Measurement Model
After conducting CFA for each factor in the first step of measurement model
testing, the CFA is performed for all the factors at the same time in step 2. In this step,
156
individual scale items were loaded on their appropriate factors and all factors were
correlated with each other.
CFA was conducted for the measurement model that is comprised of 11
factors measured by 48 scale items. The results of CFA for the overall measurement
model are shown in Table 5.42.
Table 5.42 CFA Results for Overall Measurement Model
No. of Items
2 (df) 2/df CFI GFI RMSEA Remarks
Initial Model Results
48 4887.590(1019) 4.796 0.813** 0.739*** 0.079**
Items SS1, SQ1, AR3,
SL5, PL4 and AR1 are
eliminated in modification
Refined Model Results
42 3337.989(761) 4.386 0.854** 0.796*** 0.075**
Final Refined Model Results
42 2852.743(681) 4.189 0.869** 0.808** 0.072**
* Indicates fit is good ** indicates fit is acceptable *** indicates fit is unacceptable
The initial model values of CFI and RMSEA indicate that the measurement
model fit is acceptable but the value of GFI show unacceptable fit of the measurement
model. Hence, measurement model improvement is required. For improving the
model, the items SS1, SQ1, AR3, SL5, PL4 and AR1 are eliminated. The SRW value
of AR1 (0.367) is below the cut off value of 0.4 and SRW for PL4 (0.423) is lowest of
the accepted values. The SRC values associated with the items SS1, SQ1, AR3 and
SL5 are very high and unacceptable. So, items SS1, SQ1, SQ4, SL2, PL4 and AR1
are eliminated and CFA was conducted for getting the refined model results.
In the refined model, fit indices, CFI and RMSEA, indicate that the
measurement model fit is acceptable but the value of GFI show unacceptable fit.
Hence, further model improvement is required. For improving further, six pairs of
items were correlated; AL6 and AL7, AL4 and AL6, AL2 and AL5, MQ2 and MQ5,
157
PP2 and PP4, and PP3 and PP4. These pairs of items are correlated because the SRCs
between these items are high and unacceptable. After these modifications, CFA was
conducted again. The final refined model CFA results are shown in Table 5.42.
The values of CFI, GFI and RMSEA of the final refined model show
acceptable fit. This final refined model is taken as the structural model of the study for
testing the devised hypotheses. The path diagram of the final refined measurement
model is shown in Figure 5.12.
158
Figure 5.12 Path Diagram of Final Refined Measurement Model