Discussion of Findings

The analysis results in last chapter will be discussed in the following sections from various dimensions including the quality of the questionnaire items, the result of a single path or the overall model, comparison between the models and comparison across the different extensions.

In hindsight of the result of structural model (Figure 4-1), it can be seen that the reason why the construct, perceived tie, in the model had insignificant effect might be that the quality of the measurements of the construct was poor. First of all, although all the items measuring perceived tie had pass through the criterion of the individual item reliability (0.7 of each corresponding item loading) and convergent validity (AVE > 0.5;

Cronbach’s alpha > 0.6), but the values just got a low pass (TIE3 loading = 0.701; AVE

= 0.591; Cronbach’s alpha = 0.657). In addition, there is also a research indicated that the criterion of Cronbach’s alpha should be 0.7(Anderson & Gerbing, 1988), and therefore causing the result of examining the convergent validity of the construct, perceived tie, to be failed. Lastly, despite that the obvious non-serious respondents had already been deleted, answering the reverse-coded items, TIE1 and TIE2, measuring the construct, perceived tie may still be difficult to some of the respondents. They may not

58

notice that the questions were reverse-coded or may be confused with the logic of how to answer the question. Also, after answering the question of TIE1 and TIE2, they may be mistaken that TIE3 was also a reverse-coded question and therefore answer it in the wrong way. Overall, the insignificance of the effect producing by perceived tie may cause by the short of the measuring items’ quality.

Through the result of Structural Model and analysis using the original model, the following findings were discovered. The R² value of the constructs indicated that the related results for the construct, behavior intention toward extension (0.680), had practical value, and for the construct, perceived quality of the extended app (0.227) and perceived usefulness of the extended app (0.281), had moderate explanatory power, whereas the related results for perceived tie and perceived fit had weak explanatory power. In addition, Table 4-12 Direct and indirect effects between constructs indicated that the total effect of perceived usefulness of parent app (0.036) is stronger than perceived quality of parent app (0.016) on behavior intention toward extension, which implied that the influence of perceived usefulness cannot be ignored. Also, the total effect of perceived fit on behavior intention toward extension (0.084) was considerable.

Compared with the original model proposed by (Song et al., 2010) with the same data, where the R² value of the constricts were 0.581 for behavior intention toward extension,

59

0.058 for perceived fit and perceived tie, the overall explanatory power of the proposed model in this research was stronger. However, in both model with the same data, the related effects of perceived tie remained insignificant.

In comparing the different result when using different type of extended app (Section 4.4), the outcome leads to several findings. First of all, the overall explanatory power when using LINE Camera as extended app is stronger than using LINE POP as extended app. When using LINE Camera as extended app, the R² values of behavior intention toward extension, perceived quality of extended app, perceived usefulness of extended app, perceived fit, and perceived tie were 0.695, 0.248, 0.319, 0.083 and 0.061 correspondingly, while the R² values were 0.680, 0.211, 0.259, 0.050 and 0.057 respectively when using LINE POP as extended app. Secondly, it is more likely that perceived quality of parent app through the indirect effect of perceived fit to influence perceived quality of extended app, perceived usefulness of extended app, and behavior intention toward extension when using LINE Camera as extended app, while there is more possibility that perceived usefulness of parent app through the indirect effect of perceived fit to influence the same constructs mentioned. However, in both situations the total effect of perceived usefulness of parent app on behavior intention toward extension was stronger than perceived quality of parent app. According to the research

60

report (TWNIC, 2012a), LINE messenger belongs to the app type of sociality, LINE Camera belongs to photo editing or photography, and LINE POP belongs to game type of apps. Therefore, it can be assumed that when the sociality type of apps are as the parent app, the circumstance will be more fit to this model when the extended app is a photo editing or photography app. Since both sociality and photo editing or photography types of apps perform general task while gaming type of apps perform gaming task (Fang et al., 2006), users may perceived LINE messenger and LINE Camera to be in the same category and thus more fit to the research assumption. In addition, perceived quality is more likely to transfer to the extension through perceived fit when photo editing or photography types of apps was as extension while perceived usefulness, which can be explained as perceived playfulness in gaming type apps, is more likely to transfer to gaming type extension through perceived fit. Nevertheless, the adoption intention of both types of apps would be more influenced by the perceived usefulness of their parent app than perceived quality.