Page 151 - 33-3

NTU Management Review Vol. 33 No. 3 Dec. 2023

have placed surveys on an online platform without any compensations.
Through the procedure of the questionnaire survey, we collect the data of 228
participants. After removing unqualified subjects who had no experience using health
apps but still filled out the questionnaire, 218 subjects are eligible for data analysis. Table

2 shows the detailed demographic characteristics of the eligible participants. The gender
distribution of participants is 33% men and 67% women. As for age, most participants
(60.8%) are about 20 to 29 years old. In addition, most participants have received higher

education; 40.1% of participants are college students, and 40.6% are master’s students or
above. The most popular health apps used are Apple Health (59.6%), followed by Samsung
Health (11%), Google Fit (9.6%), and MyFitnessPal (5.5%). Most participants start using
the health apps within the last 3 years (93.5%); consequently, we can observe that the use
of health apps has gradually become a trend in recent years. However, most participants

use the health apps infrequently; 86.2% of participants indicate that their total usage in the
last month is less than 10 hours, and 77.5% of participants say they used the apps less than
ten times in the last month. Generally speaking, we believe that this information will help

suppliers to understand more about what users are thinking.

5. Data Analysis and Results

This study analyzes the data collected from the questionnaire by using statistical

software SPSS 22.0 and partial least squares (PLS), which are analytical tools to validate
the research hypotheses. This study also applies structural equation modeling (SEM), a
statistical method combining factor analysis and path analysis. Among the methods of
SEM, this study uses partial least squares (PLS), and many scholars have proved that the

PLS method is one of the best methods for testing empirical SEM (Xu, Dinev, Smith,
and Hart, 2011). In order to use PLS, the number of samples must meet their standards.
The minimum sample size of PLS is ten times the number of indicators and is associated
with the most complex or largest number of endogenous constructs (Marcoulides, 1998).

The research model involves 6 constructs; therefore, the sample size of 218 meets the
requirement of the PLS. We then use Smart-PLS 2.0.M3 to analyze data. This study also
uses a two-step approach, provided by Anderson and Gerbing (1988), to analyze data
collection for this survey. The first step is to check the measurement model and the second

143