*3.3. Data and Statistical Analysis*

The partial least squares (PLS)-based path model is adopted to assess the ICFs impacting individuals' WAPP. A Likert scale consisting of five-points included 5 = "Totally agree", 4 = "Agree", 3 = "Neutral", 2 = "Disagree", and 1 = "Totally disagree." The schematic outline of the research methodology is presented in Figure 2.

**Figure 2.** Schematic outline of the research methodology. Source: Authors' elaboration.

#### 3.3.1. Demographic Data

Data on the demographic characteristics of the respondents are reported in Table 1. The participation of males (66.5%) was higher than that of females (33.4%). The proportion of urban respondents (59.3%) exceeded that of rural respondents (40.7%). The main proportion of respondents (54.7%) consisted of youth (up to 25 years old), while middle-aged individuals (26–50 years) made the second-largest age group (31.3%). The mean of respondents' age was 30.26 years, while its standard deviation was noted as 12.86. The respondents varied from illiterate (zero schooling years) to postgraduate (18 and above schooling years) in qualification. Bachelors (14 schooling years) made the largest proportion (20.9%), followed by the secondary (10 schooling years) and the higher secondary (12 schooling years) groups. The smallest proportion (4.2%) was based on illiterate respondents (zero schooling years). The largest proportion of respondents (56.6%) was unmarried, while a tiny proportion (2%) was divorced. The majority of respondents (34.2%) were employees in both public and private sectors, while students comprised the next significant share (31.3%). However, labor contributed to the smallest proportion (14.6%). The highest percentage of the respondents (43.4%) were from households with upper middle income (300,001–600,000 PRK per annum), while the lowest income households were in the smallest proportion (5.4%).


**Table 1.** Attributive profiles of the respondents.

#### 3.3.2. Statistical Measurement Model

Confirmatory factor analysis was carried out to explore whether the models were reliable and valid. The assessment of external loadings was conducted and is shown in Table 2. The external loading equivalent to or greater than 0.7 was argued to determine variations roughly surpassing 50% [97], showing that the calculated factor attained a permissible degree of reliability. As a result, external loading values above 0.7 suggest the non-exclusion of the loading factor [98].

Moreover, [99] suggested that non-external consistencies depict the reliability of a construct. In this respect, ρ-A, Cronbach-alpha (C-α), and composite reliability (CR) were employed. The range of values from 0.7 through 0.95 suggests satisfactory reliability [100]. Since C-α may understate a finite sample's efficiency, the use of an additional CR measuring tool is encouraged [101]. Furthermore, the magnitudes of ρ-A in a range between CR and Cronbach-alpha are taken to be accurate [102]. The average variance extracted (AVE) is reported in Table 2. Hair et al. [103] suggested that AVE surpassing 0.5 can be considered reliable, which is true in the present case. Thereby, the constructs in Table 2 are reliable. These findings authenticated the convergent validity and reliability of our measurement model.


**Table 2.** Measurement model results.

Notes: Degree to agree with the affirmative response is classified as: 5 = "Totally agree", 4 = "Agree", 3 = "Neutral", 2 = "Disagree", 1 = "Totally disagree." C-α: Cronbach-alpha. MAP: Mythical attitude towards pandemic, PK: Pandemic knowledge, EPPA: Ease of pandemic prevention adoption, SEF: Self-efficacy, PGB: Peer groups' beliefs, MV: Moral values, RAB: Risk-averse behavior, PR: Perceived risk, LTPW: Lack of trust in political will, WAPP: Willingness to adopt pandemic prevention. AVE: average variance extracted, CR: composite reliability, ρ-A: internal consistency reliability, C-α: Cronbach-alpha.
