*3.1. Data Collection*

Data was collected through an online survey because people today, especially the young generation, tend to have their own smartphone and spend most of their time on it. Before the survey was conducted, researchers had developed a possible list of questions based on the research topic and evaluated each question together to determine the top 10 questions to ask the participants through Google survey form and shared via social network. To achieve an explicit and superior result, questions that consist of multiple choices, including both Indonesian Bahasa and English subtitles, had been planned so that they could obtain a better understanding and easy answering for major participants. The total number of respondents in the survey is 100 students, selected at random from different academic years. Furthermore, researchers also collected secondary data from publishing websites for reference and guidance purposes only.

#### *3.2. Sampling Technique*

This study also used a non-probability sampling technique, because not all samples have criteria that are in accordance with what the authors have specified. The type of sampling used by the author is accidental sampling. This study was measured using a Likert scale by determining the level of their agreeable answers to the questions raised. The questionnaire in this study provides 1–5 scale options as an alternative answer that will be used by respondents.

#### *3.3. Data Processing*

Furthermore, the data processing was conducted using Lisrel 8.80. To clarify the variables studied as formulated in the description above, the main problems studied are product knowledge (X1), perceived benefits (X2), risk perception (X3), and usage decisions (Y). The identification of the SEM model was done by calculating the degree of freedom from the SEM model. The value of the degree of freedom from the structural equation model (SEM) plays an important role in determining whether the process to estimate the parameters of the SEM can be carried out or not. The calculation of the value of degrees of freedom from the SEM is intended to find out whether each value of the estimated parameter of the SEM has a unique value/single solution. Based on the output of Lisrel, it is known that the degree of freedom from the SEM is 38. Since the degree of freedom is greater than 0, then the process for estimating the parameters of the SEM can be carried out. Such an SEM is called an over-identified model.

Testing the inner model will give the results of the relationship between constructs. Table 1 is the result of bootstrapping, which describes the estimation results of each 5% significance.


**Table 1.** Construct Reliability and Validity.

Based on Figure 4, it is known that the latent variables in this study are product knowledge, perceived benefits, and perceived risk, which are exogenous latent variables. Meanwhile, the latent variable is decision, which is particularly an endogenous latent variable. The endogenous latent variable is characterized by an arrow that goes toward the latent variable. In other words, endogenous latent variables are influenced by other latent variables.

**Figure 4.** Structural equation model (SEM) diagram.
