The dataset was developed to support the analysis of food safety knowledge, attitudes, and practices among street food vendors in the Mpohor-Fiase District of the Western Region of Ghana. It consists of data collected from 239 respondents, corresponding to the final sample size determined through statistical sample size estimation and adjusted for non-response. The dataset is structured to align directly with the study objectives and the Knowledge–Attitude–Practice (KAP) framework underpinning the research.
The dataset is organised into four main sections. The first section contains socio-demographic variables, including gender, age group, level of education, and years of vending experience. These variables are coded numerically and measured at nominal or ordinal levels, allowing for group comparisons and stratified analyses. Additional contextual variables, such as access to clean water and prior food safety training, are included to support bivariate and multivariate analyses of food safety practices.
The second section captures food safety knowledge using seven binary items (correct/incorrect). These items assess respondents’ understanding of key food safety concepts such as foodborne illnesses, handwashing, temperature control, cross-contamination, and safe food storage. A composite KnowledgeScore variable was generated by summing responses across the seven items, producing a continuous score ranging from 0 to 7.
The third section measures food safety attitudes using seven Likert-scale items ranging from strongly disagree to strongly agree. These items assess perceptions of food safety importance, confidence in applying safety measures, views on regulatory effectiveness, and willingness to participate in training. Responses were aggregated to form a continuous Attitude Score, with higher scores indicating more positive attitudes toward food safety.
The fourth section focuses on food handling and hygiene practices, assessed through seven binary practice items reflecting compliance with recommended food safety behaviors. A composite PracticeScore was computed and further recoded into a binary outcome variable (GoodPractice) to facilitate logistic regression analysis.
