*2.2. Data Collection and Sampling*

This study adopted a mixed-method approach, combining an online questionnaire survey, semi-structured interviews (through meeting application or phone call) and participatory observation, adjusting to pandemic situations. The purposive sampling strategy was used throughout the data collection process (except for observation).

Before the actual conduct, pre-testing was carried out to examine the content validity of the questions asked in the survey (including semi-structured interview). All questions were refined based on the feedback from both experts in the field and laymen and were adapted from studies in past literature [30,33–39] to improve their reliability and representativeness to the study. The questionnaire survey was constructed using an online survey administration application, Google Form, and distributed through various online platforms (e.g., Facebook, Instagram, and WhatsApp, popular applications in Malaysia). The questions were written in three major languages: Malay, Chinese, and English. Respondents were filtered carefully under two conditions: a person (1) aged 18 years old and above who (2) has lived in Manjung district for more than a week. It was also mentioned that each

respondent would be representing a household. The questionnaire consisted of six main sections: (a) respondent's profile, (b) knowledge, (c) attitude, (d) behaviour, (e) psychology behind R&SS practices, and (f) invitation to interview. The targeted sample size for the survey was 382 respondents for 58,186 house premises (data from Manjung Municipal Council) based on the calculation by [40]; however, it was not achieved due, in part, to the relatively less effective channel of survey distribution via online networks. Only 100 respondents took part in the survey within the scheduled time frame for quantitative data collection. The mixed methods of this study underwent a sequence arrangement [41], in which the qualitative data that followed were collected after quantitative data were completed.

Seven respondents, who had answered the survey, were recruited to a semi-structured interview based on their consent and willingness to contribute to this study. This further strengthened the perspective of service users through descriptive and abstract primary data on their waste management process [42]. Questions to them were focused on the advantages and disadvantages of the R&SS, regardless of their actual participation. On the other hand, the semi-structured interview included the perspectives of the service providers to explain the condition and quality of R&SS practices provided, as well as how the households' needs were addressed. Two personnel from the local authority (the head of recycling project implementation) and non-governmental organisations (which are actively involved in R&SS implementation at household level) were interviewed. Questions for service providers merely focused on the mechanisms involved in providing their respective R&SS services and how to engage with the community.

Observation was conducted to inspect the current waste management mechanism and environment of the study area, from waste generation, storage, collection to disposal. Taking into consideration that presence of the observers might have affected the behaviour of the observed group, participatory observation was opted to blend in the situation [43]. Observation took place in public areas, including wet markets and neighbourhood streets, when municipal waste collection was in action. The observed objective focused on how humans carry out waste disposal, rather than on the humans themselves. Therefore, no consent was needed, and no personal communication was conducted. No photos were taken of any individual present at the observation spot.

#### *2.3. Data Analysis*

Since the data collected consisted of both quantitative and qualitative data, appropriate analysis techniques were used for each type of data accordingly.

Microsoft Excel 2019 was used to perform all statistical (quantitative) analysis, ranging from descriptive, inferential statistics to correlational analysis. Measures of central tendency and dispersion (mean and standard deviation) were used to study data distribution. In addition, the Chi-Square test, together with Pearson's Correlation Coefficient, which was performed at 95% confidence level, were used to determine the association between variables (two forms: nominal and continuous data). A scoring system was developed to assess the adequacy level of the KAB through a fixed range of scores modified according to the widely adopted Bloom's Cut-Off Point [38,44]. This method allowed the authors to convert different types of data, including nominal—true or false answer and percentages, and ordinal data—Likert scales, into scores—continuous data. The scores (by level) obtained were used to test the correlation between KAB components and other variables.

For knowledge (first component), 20 questions were asked in two parts: ten true or false questions (general knowledge assessment) and ten multiple choice questions (specific knowledge assessment). A score of 1 was given to each correct answer (for each sample); in contrast, no score was deducted for a wrong answer, instead it was given a score of 0. Hence, each part had a maximum and minimum score of 10 and 0, respectively. These scores were divided into three levels through a modified Bloom's Cut-Off Point, namely (a) High for 10 to 8 scores, (b) Moderate for 7 to 5 scores, and (c) Low for scores less than 5. For attitude (second component), there were 14 questions with a 5-Point Likert scale to assess the degree of importance and agreement. Scores 1 to 5 were given to respective points

on the Likert scale (in the order from Strongly Disagree to Strongly Agree). This resulted in a maximum score of 70 and a minimum of 14 for each sample. Modified Bloom's Cut-Off Point was also used to establish three levels, namely (a) Positive (70 to 52), (b) Neutral (51 to 33), and (c) Negative (14 to 32), which were equally divided and given the same class interval. Finally, for behaviour (third component), there were 10 questions with multiple choice of answers. Similar to the first component, the scores were divided into three levels through a modified Bloom's Cut Off Point, namely (a) Good (10 to 8), (b) Satisfactory (7 to 5), and (c) Poor (4 to 0).

Phenomenological (qualitative) analysis was used to carefully record the conduct of waste management by the service users during the observation and semi-structured interview processes, respectively. This was to ensure that both observed and elicited data could be precisely analysed and transformed into useful information for the description of a phenomenon. Additionally, it could extract the perception of the service providers on the outcomes of the waste management service [45].
