Understanding consumer online shopping behaviour from the perspective of transaction costs



Download 1,58 Mb.
Pdf ko'rish
bet95/214
Sana11.07.2022
Hajmi1,58 Mb.
#773924
1   ...   91   92   93   94   95   96   97   98   ...   214
Bog'liq
Gao whole thesis

4.9.4 Model Estimation 
The most widely used type of estimation is maximum likelihood (ML), followed by 
generalized least square (GLS) and unweighted least squares (ULS) (Anderson and Gerbing 
1988, Bollen 1989, Hoyle 1995, Kelloway 1996). The ML estimates are consistent, unbiased, 
efficient, scale-invariant and scale-free under the assumption of normality (Schumacker and 
Lomax 2004). As indicated by SEM scholars, the ML estimates are highly appropriate when 
the observed variables are interval scaled and normal distributed (or minor deviations). 
Nevertheless, if observed variables are ordinal scaled and/or extremely skewed or peaked 
(non-normally distributed), then the ML estimates are not robust (Anderson and Gerbing 
1988, Bollen 1989, Hoyle 1995, Kelloway 1996). The GLS estimates have the same 
properties as the ML approach under a less stringent normality assumption. Meanwhile, the 
ULS estimates do not depend on a normality distribution assumption, but the estimates are 
neither efficient, scale-invariant nor scale-free (Schumacker and Lomax 2004). In this study, 
the ML estimation method was used to estimate the parameters in the proposed hypothesized 
models. The current data were screened and examined for normality prior to their analysis in 
AMOS 21.0 (see the next chapter).
4.9.5 Model Fit Testing 
A number of goodness-of-fit measures have been developed to assist in interpreting how well 
the structural equation models fit the sample data. Such measures can be classified into 
absolute fit, incremental fit and parsimonious fit measures (Ho 2006, Hair
 et al.
2006). As 
described by McDonald and Ho (1999), absolute fit measures determine how well the a priori 
model fits, or reproduces the data. Some commonly used measures of absolute fit include the 
chi-square (
χ
2
), the goodness-of-fit index (GFI), the root mean square error of approximation 


200
(RMSEA), and the standardized root mean squared residual (SRMR). Incremental fit 
measures compare the proposed model to a null model to determine the degree of 
improvement over the null model. Common incremental fit measures are the comparative fit 
index (CFI), the normed fit index (NFI) and the Tucker-Lewis index (TLI). Parsimonious fit 
measures evaluate the fit of the model versus the number of estimated coefficients needed to 
achieve that level of fit. Examples of these measures include the parsimonious normed fit 
index (PNFI) and the parsimonious goodness-of-fit index (PGFI) (Tian and Stewart 2007). 
Given a lack of consensus on the best measure of fit, SEM scholars recommend the use of 
multiple measures rather than rely on a single choice in evaluating the model fit (Bollen 1989, 
Hu and Bentler 1999, Schumacker and Lomax 2004, Ho 2006, Byrne 2009, Hair
 et al.
2006). 
As indicated by Byrne (2009), the chi-square (
χ
2
) is the only statistical test of significance for 
testing the fit of a proposed model. A low 
χ
2
value relative to the degrees of freedom (
df
), 
indicating non-significance (p > 0.05), would point to a good fit. This is because such non-
significance means that there is no statistical difference between the actual and predicted 
input matrices (Byrne 2009). Nevertheless, there are two cases in which the 
χ

may be 
misleading (Hoyle 1995). Firstly, the 
χ

is very sensitive to the complexity of the model; the 
more complex the model, the more likely the results will indicate a poor model fit. Secondly, 
it is too sensitive to the sample size; the larger the sample size, the more likely it is that the 
model will be rejected even if it is, in reality, a good fit with the data (a Type II error- 
accepting a false null hypothesis).
To recognize these problems, this study complements the 
χ

measure with other goodness-of-fit measures. 
Marsh et al. (1988) propose that the criteria for ideal fit measures are relative independence 
of sample size, accuracy and consistency to assess different models, and ease of interpretation 


201
aided by a well-defined pre-set range. Based on these stated criteria, Grace and Weaven 
(2011) recommend the use of the GFI, CFI, TLI and RMSEA. Also, in their studies of the 
performance of various fit measures in relation to sensitivity to model misspecification, Hu 
and Bentler (2010) and Kline (2012) recommend the use of the RMR in tandem with one of 
several other indices particularly the TLI and RMSEA. In line with these recommendations, 
the present study reports the GFI, CFI, TLI, RMSEA and RMR in combination with reporting 
the 
χ
2
. The following is a brief discussion of the fit measures and their interpretation taken 
into account by the present study when testing model fit. 
The GFI is a measure of fit between the hypothesized model and the observed covariance 
matrix. Holmes-Smith and Coote (2012) argue that the good model fit is achieved if GFI is 
above 0.90. The CFI estimates the proportion of improvement in the proposed model beyond 
the null model, based on the noncentral 
χ
2
distribution that allows for unbiased estimation of 
small sample size. The CFI has a value ranging from 0 to 1 (1 = perfect fit), with a value of 
0.90 or greater representing a good-fitting model. Similarly, the TLI indicates the percentage 
improvement in fit over the null model by taking into account 
df
. Higher values of the TLI 
indicate a better fitting model, and it is common to apply the 0.90 rule as indicating a good fit 
to the data. RMSEA estimates the lack of fit in a model compared to a perfect (saturated) 
model. A value of 0.08 or less indicates a good-fitting model, while a value greater than 0.1 
represents a poor-fitting model (Steiger 1990, Browne and Cudeck 1993). The RMR is the 
square root of the discrepancy between the sample covariance matrix and the model 
covariance matrix. The RMR values range from 0 to 1, with value of 0.05 or less indicating a 
good model fit (Hooper
 et al.
2008).


202
Importantly, the statistical significance of individual parameter estimates for the paths in the 
proposed model also needs to be considered for assessing model fit (Schumacker and Lomax 
2004). A ratio of the parameter estimate to the estimated standard error can be formed as a 
critical value. If the critical ratio exceeds 1.96 at a significant level of less than 0.05, then the 
parameter is statistically significant. Furthermore, as suggested by Schumacker and Lomax 
(2004), the parameter estimates obtained from the analysis should be theoretically meaningful
and should be within an expected range of values (e.g., variances should not have negative 
values and correlations should not exceed 1). 

Download 1,58 Mb.

Do'stlaringiz bilan baham:
1   ...   91   92   93   94   95   96   97   98   ...   214




Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©hozir.org 2024
ma'muriyatiga murojaat qiling

kiriting | ro'yxatdan o'tish
    Bosh sahifa
юртда тантана
Боғда битган
Бугун юртда
Эшитганлар жилманглар
Эшитмадим деманглар
битган бодомлар
Yangiariq tumani
qitish marakazi
Raqamli texnologiyalar
ilishida muhokamadan
tasdiqqa tavsiya
tavsiya etilgan
iqtisodiyot kafedrasi
steiermarkischen landesregierung
asarlaringizni yuboring
o'zingizning asarlaringizni
Iltimos faqat
faqat o'zingizning
steierm rkischen
landesregierung fachabteilung
rkischen landesregierung
hamshira loyihasi
loyihasi mavsum
faolyatining oqibatlari
asosiy adabiyotlar
fakulteti ahborot
ahborot havfsizligi
havfsizligi kafedrasi
fanidan bo’yicha
fakulteti iqtisodiyot
boshqaruv fakulteti
chiqarishda boshqaruv
ishlab chiqarishda
iqtisodiyot fakultet
multiservis tarmoqlari
fanidan asosiy
Uzbek fanidan
mavzulari potok
asosidagi multiservis
'aliyyil a'ziym
billahil 'aliyyil
illaa billahil
quvvata illaa
falah' deganida
Kompyuter savodxonligi
bo’yicha mustaqil
'alal falah'
Hayya 'alal
'alas soloh
Hayya 'alas
mavsum boyicha


yuklab olish