Demographic Statistic

After organizing the data, 506 samples were assessed for three major biases, which are noncoverage bias, sampling bias and nonresponse bias, in online survey (Best, Krueger, Hubbard, & Smith, 2001; Couper, 2000), to ensure the representativeness of the samples.

First, noncoverage bias means that not everyone in the research target population is in the sample frame, which in this research it represents that not everyone who use mobile app can access the online survey which distributed in the web forums and Facebook. A report in 2012 (TWNIC, 2012b) indicated that there is more than 77.25%

of people in Taiwan who has the experience in surfing internet, therefore the high internet penetration imply that whether the potential respondents have access to the internet so that they can participate in the online survey is not a serious problem.

Additionally, the distributed channels, web forums and Facebook are popular in Taiwan and open to everyone on the internet. Thus the research has no serious noncoverage bias.

Secondly, sampling bias refers to the situation that not all members of the sample frame are measured in the research. Because this research did not apply random

36

sampling method, there is potential of sampling bias. In order to judge sampling method, the demographic of the research samples are compared to the demographic of Taiwan smart device users¹ investigated by (Google Inc., 2012) in the website, Our Mobile Planet, indicates the gender distribution of the research samples, which is similar to the findings of Our Mobile Planet, where the distribution is male for 55% and female for 45%. Furthermore, a t test was performed, and the result (p=0.345 > 0.05) implies that there is no significant difference between these gender distributions. However, the education level and age distribution (Table 4-1) of the samples were centralize at higher education level and younger age, which mostly are college and graduate students. This is probably resulted of that the user of the web forums and Facebook who often participate in answering questionnaires are generally young people and high educational students. Nevertheless, the sample still considered representative for following reasons.

According to the summary of the investigation report of TDCDA (2012), smart device users, 20-29 years old accounted for 83.9% and university or higher education level accounted for 83.2%, have higher proportion of downloading app, which is in accordance with the samples distribution of this research samples. Besides, in prior study (Walczuch & Lundgren, 2004), the feasibility of using student sample in e-tailing

1 According to (TWNIC, 2012a), there is 66.92% of smart device users who has downloaded apps.

37

researches is confirmed, and still many researches (Kasesniemi & Rautiainen, 2002;

Katz & Sugiyama, 2006; Ling & Yttri, 2002; Ozok & Wei, 2010; Skog, 2002) demonstrate the use of student samples in research of mobile field, therefore the use of student samples in mobile app adoption are feasible. Also, the high educational level of the samples implies that the respondents were more likely to understand the survey items and thus present more reasonable responses. Overall, the sampling bias is not a serious problem in this research.

Table 4-1 Demographic distribution of sample respondents

Gender Education Age

Male Female

Under High school

College

Above Master

degree

<18 18-24 25-30 >30

49.40% 50.60% 3.20% 61.30% 35.60% 0.80% 56.50% 30.80% 11.90%

Finally the nonresponse bias was assessed by comparing the early respondents with later respondents, assuming that later respondents was more similar to nonrespondents because they were less readily to respond and might response owing to the increasing call for participating the survey (Armstrong, Armstrong, & Overton, 1977).

Nonresponse bias means that the answers of respondents are different from the answers of those who did not answer, which is a critical problem occurs in online survey (Best et al., 2001; Couper, 2000; Gosling et al., 2004). The demographic distribution of early

38

and late respondents are shown in Table 4-2, Table 4-3 and Table 4-4, and independent-samples t-test was conducted to see if there is any difference of early and late respondents in gender, education level and age distribution. The test result of all three distribution (p>0.05) indicated that the two groups of respondents were no significantly different in the demographic distribution. As a result, the nonresponse bias is not likely to occur.

Table 4-2 Gender distribution of early and late respondents Gender

Male Female

Early respondents 50.80% 49.20%

Late respondents 48.00% 52.00%

p value 0.662

Table 4-3 Education level distribution of early and late respondents Education

Early respondents 3.20% 57.90% 38.90%

Late respondents 3.10% 64.60% 32.30%

p value 0.327

Table 4-4 Age distribution of early and late respondents Age

<18 18-24 25-30 >30

Early respondents 0% 56.30% 31.70% 11.90%

Late respondents 1.60% 56.70% 29.90% 11.80%

p value 0.689

39

In addition, the proportion of the users who have used the parent app, LINE messenger, and the extended app, LINE Camera or LINE POP, are list below (Table 4-5). It can be seen that most of the users in the research sample have used the parent app, more than half of them have used the extended app, less than one third of them haven’t used both of the parent app and the extended app, which indicated that the sample user have sufficient capability to answer related questions of the apps.

Table 4-5 Usage rate of the parent app and the extended app

usage rate Parent app Extended app Both the parent app and the extended app

yes 97.63% 55.53% 71.94%

no 2.37% 44.47% 28.06%