Page 75 - Azerbaijan State University of Economics
P. 75

Maral Jamalova: Socıal Influence and Smartphone Adoptıon in Collectıvıstıc Country:
                                                        Evıdence from Azerbaıjan


                    Additionally, statistical literature confirms that reliability estimates are also used as
                    evidence  of  convergent  validity  (Joseph  F.  Jr  Hair  et  al.,  2014,  p.  619)  and  high
                    scores of average variance extracted (AVE) prove convergent validity. In this case,
                    AVE  is  the  share  of  total  variance  explained  by  the  latent  variable  (Malhotra  &
                    Birks, 2010, p. 734). The accepted level of  AVE  (Joseph F. Jr Hair et  al.,  2014;
                    Tabachnick & Fidell, 2013) is greater than 0.5 (see Table 2). Except for Facilitating
                    Conditions that were deleted from the study, the results of AVE for the remaining
                    constructs were in the accepted range.

                    Table 2: Convergent validity and reliability
                                                              Composite      Average Variance
                       Indicators      Cronbach's Alpha
                                                              Reliability    Extracted (AVE)
                        BA Az*             two items            0.827             0.711
                        HM Az                0.884              0.927             0.81
                        HT Az*             two items            0.682             0.521
                        PV Az*             two items            0.806             0.681
                        SBI Az*            two items            0.825             0.703
                        SI Az*             two items            0.73              0.577

                    Note: BA – Brand Awareness; SBI- Symbolic Brand Image; HM – Hedonic
                    Motivation; HT – Habit; PV – Price Value; SI – Social Influence; USE – Actual
                    Usage.
                    Note: *- Spearman’s correlation coefficient was calculated instead of Cronbach’s
                    Alpha.
                    Source: own editing based on the results of the analysis

                    The other important criterion for ensuring discriminant validity is called the Fornell-
                    Larcker  criterion.  The  results  for  the  criterion  fulfill  the  main  requirement  which
                    states  that  “the  square  root  of  each  construct’s  AVE  should  be  greater  than  its
                    highest correlation with any other construct” (Joe F. Jr Hair, Sarstedt, Hopkins, &
                    Kuppelwieser, 2014, p. 139).

                    Considering that actual use is a formative construct, rules for formative measurement
                    indicators were applied. It must be reported that there was no multi-collinearity issue
                    (VIF<3).  Based  on  the  recommendation  of  Hair  et  al.  (2017,  p.  170)  the  items
                    illustrating outer loading higher than 0.5 (i.e., U5 and U7) were kept for further analysis.










                                                           75
   70   71   72   73   74   75   76   77   78   79   80