This chapter presents the empirical consequences of the studies which were obtained by carry oning several statistical analyses. First, a demographic analysis of the samples is presented on footing of the gender and the age of the respondents. The following subdivision is dedicated to the analysis of the internal consistence dependability of the multiple-item graduated tables, followed by a factor analysis. Descriptive statistics every bit good as a correlativity analysis are so presented. All these old analyses were performed thanks to package called SPSS Statistics 21.Finally, the measuring theoretical account and the hypotheses are tested with package called SmartPLS.
As mentioned in the old chapter, two types of Facebook trade name communities were examined ; hence, the empirical research was conducted on two samples of Facebook users. The first sample is made up of members of marketer-initiated trade name communities ( Official Fan Pages ) and the 2nd one consists of members of consumer-initiated trade name communities ( Groups ) . However, it is of import to state that the informations were non analyzed in two separate parts and therefore, merely one database was created. The type of trade name community is tested as chairing variable in the last subdivision of this chapter.
There's a specialist from your university waiting to help you with that essay.
Tell us what you need to have done now!
order now
A sum of 265 replies were gathered: 128 from the first selected sample and 137 from the 2nd selected sample. However, out of this sum of started questionnaires, a certain sum of informations had to be dropped in order to continue to the statistical analyses with a clean database. On the one manus, outliers were detected with the analysis of point ‘s Z-scores and removed. Outliers are respondents who answered the inquiries incoherently or given the same reply throughout the questionnaire. On the other manus, merely questionnaires with at least 90 % of the inquiries answered were kept. This led to rejecting 45 replies ( 14 for sample1 and 31 for sample2 ) and obtaining a clean database of 220 responses whose all the point ‘s Z-scores were inferior to 3 in absolute value.
Demographic analysis
At the terminal of the questionnaire, participants were asked to finish two inquiries about their gender and their age, taking to pull up a sample profile. The nationality of the respondents was non requested since what mattered was that they understand easy and decently English since the questionnaire were built in this linguistic communication. The undermentioned consequences were observed:
Table 2: Respondents ‘ gender
Gender
Male
Female
Sum of respondents
TYPE_FBC_1**
Nitrogen
50
64
114
%
44 %
56 %
100 %
TYPE_FBC_2*
Nitrogen
67
39
106
%
63 %
37 %
100 %
Sum of samples
Nitrogen
117
103
220
%
53 %
47 %
100 %
**TYPE_FBC_1 corresponds to the first sample, viz. members of official Fan Pages
*TYPE_FBC_2 corresponds to the 2nd sample, viz. members of groups.
Table 2 shows that 53 % of the sum of the people who completed the questionnaires were male and 47 % were female. Sing the first sample, the proportion of work forces and adult females are rather balanced, with severally 44 % and 56 % . However, the 2nd sample shows that a higher sum of respondents were male ( 63 % ) . This observation can be partly explained by the fact that groups that were chosen for this research support trade names for which work forces are most prone to be interested in.
Table 3: Respondents ‘ age
Age
16-20
21-25
26-30
& gt ; 30
Sum of respondents
TYPE_FBC_1
Nitrogen
8
47
43
16
114
%
7 %
41 %
38 %
14 %
100 %
TYPE_FBC_2
Nitrogen
6
39
41
20
106
%
6 %
37 %
39 %
19 %
100 %
Sum of samples
Nitrogen
14
86
84
36
220
%
6 %
39 %
38 %
16 %
100 %
Sing the age of the respondents, the bulk of the sum of the participants were between 21 and 25 old ages old, followed closely by the 26-30 age bracket, with severally 39 % and 38 % . For the first sample, these consequences can partly be explained by the fact that most of the individuals contacted for the study were college pupils or immature professionals. For the 2nd sample, we can reason that immature grownups of Facebook groups are the most prone to reply such a study. Merely 16 % of the respondents are over 30 and 6 % are between 16 and 20 old ages old.
Dependability analysis: Cronbach ‘s alpha
The internal consistence dependability of the multiple-item graduated tables was tested by calculating Cronbach ‘s alpha coefficients, stand foring the averaged correlativity between the points. Specifically, Cronbach ‘s alpha can be defined as “ the norm of all possible split-half coefficients ensuing from different ways of dividing scale points ” ( Malhotra & A ; Birks, 2007, p.358 ) . This dependability index by and large varies from 0 to 1 and a value higher than 0.7 indicates satisfactory internal consistence. The closer to 1 the coefficient ‘s value, the more dependable the internal consistence of the points in the graduated table is ( J.A. Gliem & A ; R.R Gliem, 2003 ) . It is interesting to cognize that when the Cronbach alpha of a variable is below the satisfactory bound, it can be significantly improved by taking points from the related variable. The Cronbach alpha may besides be negative, which indicates a negative mean covariance among points ( Nichols, 1999 ) and hence, it is removed.
Table 4: Cronbach ‘s alpha coefficients
Items
Abbreviations
Cronbach ‘s alpha
Information quality
Intelligence quotient
0.794
System quality
SQ
0.863
Wagess for activities
REW
0.889
Interaction between members
INT
0.874
Shared values
SHVA
0.890
Community committedness
CCOM
0.831
Brand trueness
BLOY
0.806
As shown in the Table 4, all Cronbach ‘s alpha coefficients exceed 0.7 showing the internal consistence of the multiple-item graduated tables and therefore its dependability. Cronbach ‘s alpha of two variables ( Rewards for activities and shared values ) are really near to 0.9 which can be considered as an first-class internal consistence ( George & A ; Mallery, 2003, cited in Gliem, & A ; Gliem, 2003 ) .
Factor analysis
After proving the dependability of each variable of the structural theoretical account, a factor analysis was carried out in order to measure the dimensionality of the different variables. Table 5 presents the factor burdens of the points. In order to pull out the factors burdens, a chief constituent analysis with varimax rotary motion as method was made in SPSS. A burden can be interpreted as the strength of the impact between an point and the variable it measures. The higher the burden, the higher the influence of the point on the latent variable is. Hulland ( 1999 ) suggests that points with lading below 0.5 are non considered as valid and therefore should be dropped.
As shown in the tabular array 5, all point burdens were above 0.5 and therefore were considered as valid except CCOM_4 that was found to hold a non-valid burden of 0.498. The point loadings unbroken scope from 0.664 to 0.875.
Table 5: Item burdens
Items
Loads
Items
Loads
Information quality
Shared values
IQ_1
0.826
SHVA_1
0.773
IQ_2
0.858
SHVA_2
0.779
IQ_3
0.702
SHVA_3
0.770
IQ_4
0.681
System quality
Community committedness
SQ_1
0.877
CCOM_1
0.744
SQ_2
0.850
CCOM_2
0.813
SQ_3
0.742
CCOM_3
0.801
Wagess for activities
CCOM_4
.
REW_1
0.741
CCOM_5
0.664
REW_2
0.879
REW_3
0.875
Interaction between members
Brand trueness
INT_1
0.831
BLOY_1
0.771
INT_2
0.804
BLOY_2
0.828
INT_3
0.848
BLOY_3
0.705
INT_4
0.740
Descriptive statistics
Once points proved to be dependable and significantly correlated with the variable they measure were kept, we can make new variables. Each new variable was created by ciphering the mean of the related points. Therefore, the variable IQ consists of the mean of the points IQ1, IQ2, IQ3 and IQ4. The 2nd new variable – SQ – was created by ciphering the mean of SQ1, SQ2 and SQ3. The following variable REW consists of REW1, REW2 and REW3. The variable INT was built by ciphering the mean of INT1, INT2, INT3 and INT4. Then, we created SHVA which is the norm of three points ( SHVA1, SHVA2 and SHVA3 ) . The variable CCOM consists of the mean of the four points kept ( CCOM1, CCOM2, CCOM3 and CCOM5 ) , given that CCOM4 was found non valid. Finally, we built the variable BLOY which is the mean of BLOY1, BLOY2 and BLOY3.
Descriptive statistics in Table 6 present the lower limit mark, maximal mark, figure of responses, mean, standard divergence, symmetric index ( Skewness ) and flatness index ( Kurtosis ) for each variable.
Table 6: Descriptive statistics
Variables
Nitrogen
Minute
Soap
Mean
Std. Dev
Lopsidedness
Kurtosis
Intelligence quotient
220
2
7
5.32
0.79
-0.30
-0.11
SQ
220
1
7
4.35
1.08
-0.01
-0.73
REW
220
1
7
3.54
1.37
-0.09
-0.96
INT
220
2
7
5.23
0.86
-0.61
0.48
SHVA
220
1
7
4.44
1.16
-0.14
-0.89
CCOM
220
1
7
4.65
0.70
-0.21
-0.24
BLOY
220
3
7
5.86
0.72
-0.32
-0.32
Several tax write-offs can be drawn from the tabular array 6. First, it can be said that participants show trueness to the trade name supported by the Facebook community in which they are member. Indeed, the variable BLOY has obtained a minimum rate of 3 out of 7, supplying the highest mean among the variables. High agencies are besides observed by IQ and SQ. We can besides detect that the agencies of the bulk of the variables are above the mean rate of 4 on the 7-point Likert graduated table. The lone mark below the norm is obtained by the variable REW. Sing the standard divergences, they vary between 0.67 and 1.37. Finally, Skewness and Kurtosis values are all smaller than 3 in absolute value, which is the bound in order to corroborate the normalcy of the values.
Correlation analysis
Table 7 hereunder presents the correlativity matrix and shows the correlativities bing between the different variables of the theoretical account. It is of import to cognize that this statistical analysis allows detecting to what extent a variable is changing with another, but it does non let explicating the possible causality effects between them.
Correlation coefficients ( R ) can change from -1 to 1. On the one manus, a negative correlativity means that, when the variable ten additions, the variable Yttrium lessenings in the same proportion and frailty versa. On the other manus, a positive correlativity means that, the variables vary in the same manner. In add-on, the closer to -1 or 1 the correlativity coefficient, the stronger the variables are associated ; whereas a coefficient ‘s value near to 0 indicates no correlativity between the variables. The strength of the correlativity between variables can be observed thanks to the “ Balis de Cohen ” ( 1988 ) : a coefficient near to 0.1 indicates a low correlativity, near to 0.3 indicates a satisfactory correlativity and below 0.5 indicates a strong correlativity. Finally, correlativity coefficients are considered to be important at the 0.01 or 0.05 degree ( Pearson coefficient ) .
Table 7: Correlation Matrix
Intelligence quotient
SQ
REW
INT
SHVA
CCOM
BLOY
Intelligence quotient
1
SQ
,174**
1
REW
,172*
,629**
1
INT
, 197**
-,205**
-,153**
1
SHVA
,022
-,465**
-,459**
,539**
1
CCOM
,319**
,146*
,135*
,441**
,456**
1
BLOY
,342**
,082
,064
,374**
,345**
,600**
1
** Correlation is important at the 0.01 degree ( 2-tailed )
* Correlation is important at the 0.05 degree ( 2-tailed )
Correlations linked to hypotheses in respect to our theoretical account are highlighted in Grey. As we can see in Table 7, all the correlativity coefficients do non show positive important correlativity at the 0.01 degree. Indeed, the strength of the correlativities between, on the one manus, SQ and CCOM and on the other manus, between REW and CCOM are rather low with severally a coefficient ‘s value of 0.146 and 0.135. The highest correlativity is observed for CCOM and its nexus with BLOY with a coefficient of 0.600.
From the relationships between the variables, we can cipher the finding coefficients ( RA? ) . Determination coefficient identifies the proportion of discrepancy of an endogenous variable that is determined based on the variableness of the exogenic variables ( Skeskin, 2000 ) . For illustration, we can measure the finding coefficient of the nexus between SHVA and CCOM, which is equal to ( 0.456 ) A? . This means that SHVA histories for 20.8 % of the fluctuation of CCOM. Therefore, 80.2 % of the variableness of CCOM is explained by variables other than SHVA.
Model and Hypotheses proving
In order to prove the structural theoretical account and hence the related hypotheses, a package called SmartPLS utilizing Partial Least Squared as method was used. PLS way patterning – besides known as PLS Structural Equation Modeling ( PLS-SEM ) – is surely the most popular method to execute variance-based structural equation mold ( SEM ) ( Haenlein, 2004 ; Henseler, Ringle & A ; Sinkovics, 2009 ) . SME is a statistical technique used to analyze the causal links among variables of a theoretical account.
In the instance of the research, the usage of PLS was peculiarly suited for two chief grounds. First, PLS is typically recommended for research with small-sized samples, which is comparatively the instance within this thesis. Harmonizing to Goodhue, Lewis and Thompson ( 2006 ) , PLS have particular abilities that make this technique more appropriate than others when analysing little sample sizes. Second, the pick of utilizing PLS was motivated by the fact that chairing effects are analyzed within this thesis. Indeed, PLS way mold is more appropriated to complex theoretical accounts affecting hierarchal concepts, interceding and chairing effects ( Chin, Marcolin & A ; Newsted, 2003 ; cited in Wetzels, Odekerken-Schroder & A ; Van Oppen, 2009 ) .
Through PLS analysis, the way coefficient ( ? ) was computed with PLS algorithm. This coefficient can be interpreted as the comparative consequence that exists between two variables in the theoretical account. Then, the t-values were computed with the bootstrapping map and stand for the statistical significances of an independent variable in explicating a dependant variable ( Skeskin, 2000 ) . Finally, given a assurance interval of 95 % , the hypothesized relationships must obtain a t-value higher-up to 1.96 in order to be important.
Hypothesiss proving: Features of on-line trade name community committedness
In table 8, the hypotheses refering the determiners of Facebook community committedness are tested. The t-values show that three out of five hypothesized online trade name community features have a important impact on community committedness, viz. the quality of information, the interaction and the shared values. The beta coefficients indicate that their effects are positive. Sing the quality of the system ( H1-b ) and the wagess for activities ( H1-c ) , they do non accomplish important values ; hence, these hypotheses are non supported.
Table 8: Hypothesiss proving ( H1-a to H1-e )
Hypothesiss
Relationships
?
t-value
p-value
H1-a
IQ i? CCOM
0.217
4.144
& lt ; 0.05
H1-b
SQ i? CCOM
0.045
0.456
& lt ; 0.05
H1-c
REW i? CCOM
0.048
0.545
& lt ; 0.05
H1-d
INT i? CCOM
0.291
4.570
& lt ; 0.05
H1-e
SHVA i? CCOM
0.473
5.235
& lt ; 0.05
Moderating effects proving: type of Facebook trade name community
As stated antecedently in this research, we expect that the type of trade name community plays a function in our theoretical account given that they operate otherwise. Therefore, we analyzed whether the type of Facebook trade name community has chairing effects on the relationships between the trade name community features and community committedness.
As Baron and Kenny ( 1986 ) explained, three causal waies must be computed in order to prove a moderating variable[ 1 ]: 1 between the forecaster ( independent variable ) and the result ( dependent variable ) ( path a ) , one between the moderator and the result ( way B ) , and the 3rd one between the interaction term and the result ( path degree Celsius ) . The interaction term consequences from the generation of the forecaster by the moderator. Baron and Kenny ( 1986, p.1174 ) further set up that a moderator hypothesis is merely supported if path degree Celsius is important. In order to be important, the hypotheses has to hold a t-value higher than 1.96 ( 5 % , two-tailed ) . Table 9 presents the different interaction waies of the undermentioned hypotheses: H2-a to H2-e.
Table 9: Moderating effects proving: Waies c ( H2-a to H2-e )
Hypothesiss
Waies degree Celsiuss
?
t-value
p-value
H2-a
IQ x TYPE_FBC i? CCOM
-0.160
1.698
& lt ; 0.05
H2-b
SQ x TYPE_FBC i? CCOM
-0.061
1.096
& lt ; 0.05
H2-c
REW x TYPE_FBC i? CCOM
0.104
1.401
& lt ; 0.05
H2-d
INT x TYPE_FBC i? CCOM
0.171
1.542
& lt ; 0.05
H2-e
SHVA x TYPE_OBC i? CCOM
0.151
2.179
& lt ; 0.05
From what Table 9 shown, we can see that the type of Facebook trade name community has a important moderating consequence on the relationship between shared values between members ( SHVA ) and community committedness ( CCOM ) with a t-value of 2.179. Sing the other hypotheses, even if several way coefficients indicate some possible effects, none of them have important t-values. Hence, these hypotheses ( H2-a to H2-d ) are rejected.
Since the H2-e was supported, an independent-means t-test was conducted in order to compare the agencies between the two types of Facebook trade name community. In Table 10.1 here below, we can see that the mean obtained by SHVA is comparatively lower for the Official Fan Pages ( M= 3.59, SE= 0.08 ) than for the Facebook Groups ( M= 5.40, SE=0.06 ) . The 2nd tabular array indicates that the difference is statistically important T ( 218 ) = 17.136, p=0 ; reenforcing the old decision showing that the type of Facebook community has a chairing consequence on the relationship between SHVA and CCOM ( H2-e ) .
Table 10: Independent-means t-test
10.1: Groups Statisticss
TYPE_FBC
Nitrogen
Mean
Std. Dev
Std. Error Mean
SHVA
1
114
3.59
0.84
0.08
2
106
5.40
0.66
0.06
10.2: Independent Sample trial
Thymine
Df
Sig ( 2-tailed )
Equal discrepancies assumed
17.136
218
.000
Equal discrepancies non assumed
-17.190
211.6
.000
Hypothesis on the relationship between community committedness and trade name trueness
Table 11: Hypothesis testing ( H3 )
Hypothesis
Relationship
?
t-value
p-value
H3
CCOM i? BLOY
0.603
13.721
& lt ; 0.05
Table 11 hereunder nowadayss the consequences sing H3 and shows that trade name community committedness has a strong impact on trade name trueness with a t-value of 13.721. In add-on, the related beta coefficient of 0.603 demonstrates a positive relationship. Therefore, it can be said that hypothesis 3 is clearly supported.
Table 12 amounts up the assorted decisions drawn up sing the hypotheses tested in respect to our theoretical account. We can see that merely H1-a, H1-d, H1-e, H2-e and H3 are supported ; so other hypotheses are rejected.