Full dataset with the participation of respondents from São Paulo and Salvador, from September 5th to 23rd, 2019.

A consumer survey start-up (Opinion Box®) was hired to apply the questionnaire to individuals over 18 years-old, from September 5th to 23rd, 2019. The survey was elaborated based on a previous study of consumers in Southern Brazil, and composed of 15 multiple-choice and two open-ended questions. Questions 1 to 7 were related to sociodemographic data (gender, income, age, region, city, education, frequency of meat consumption). The questions from 12 to 17 presented the following response categories: very negative, negative, neutral, positive, and very positive. Questions 8 and 10 featured an image depicting conventional and cultivated meat chains.The answers to open-ended questions were analyzed by two researchers, who individually synthesized the phrases into preliminary central ideas, using the Collective Subject Discourse (CSD) method. Then, the results of the individual analyses by the two researchers were considered to define the final central ideas used, which were translated into visual representations in word clouds, where words that come up most frequently are displayed in proportionally larger sizes. The justification for answers to question 11 were also analyzed by the two different researchers, grouping them by similarity, using the CSD method.
The following factors were considered as explanatory variables: city (São Paulo, Salvador), gender (male, female), age (18 to 29, 30 to 49, 50 or older), income (up to R$1996.00, R$1997.00 to R$4990.00, more than R$4990.00), education (elementary/high school, higher education, postgraduate degree), and frequency of meat consumption (does not eat, 1 to 3 days a week, 4 to 7 days a week). Knowledge about cultivated meat, consumption intention, and the relationships between cultivated and conventional meat with the environment, human health and animal welfare were the subjects addressed in the questions. For the binary responses (yes or no) we fitted binary logistic regression models, and the results are presented through the odds ratios of a positive response and corresponding confidence intervals (95%), whereas the ordinal response variables (very negative, negative, neutral, positive, very positive) were analyzed by fitting proportional odds regression models, and the results are presented by means of the odds ratios for a more favorable (positive) response. The ordinal scale variables were grouped before fitting the regression models, by combining the two less favorable options as a negative response, and the two more favorable as a positive response. In this case, each ordinal variable presents three categories: negative, neutral, and positive. All questions contained the "I do not know" answer option. Thus, an examination was also carried out on how demographic factors influenced knowledge, or lack thereof, in relation to each of the questions. We present the non-adjusted results, based on univariate regression models, as well as the adjusted results, provided by multiple regression models. All conclusions are based on a 5% significance level. All analyses were carried out using the R software for statistical computation, version 4.3.1. The R library ordinal was used to fit the ordinal regression models.

本研究委托消费者调研初创企业Opinion Box®于2019年9月5日至23日期间，面向18周岁以上的个体开展问卷调研。本次调研问卷基于巴西南部地区消费者的既往研究设计，共包含15道单项选择题与2道开放式问题。第1至7题围绕社会人口统计学特征（性别、收入、年龄、地区、城市、受教育程度、肉类食用频率）展开。第12至17题的作答选项分为以下五类：极负面、负面、中性、正面、极正面。第8题与第10题配有展示传统肉类产业链与**培养肉（cultivated meat）**产业链的示意图。两名研究人员采用**集体主体话语法（Collective Subject Discourse, CSD）**对开放式问题的作答内容进行分析：首先分别将作答语句归纳为初步核心观点；随后结合两名研究人员的独立分析结果，确定最终采用的核心观点，并将其转化为词云可视化结果——出现频率越高的词汇，其字体尺寸比例越大。第11题的作答理由同样由两名研究人员采用集体主体话语法进行分析，按照内容相似度进行归类。本研究选取以下因素作为解释变量：城市（圣保罗、萨尔瓦多）、性别（男、女）、年龄组（18-29岁、30-49岁、50岁及以上）、收入水平（≤1996雷亚尔、1997-4990雷亚尔、＞4990雷亚尔）、受教育程度（小学/高中、高等教育、研究生学历）以及肉类食用频率（不食用、每周1-3天、每周4-7天）。问卷内容涵盖培养肉认知情况、消费意愿，以及培养肉与传统肉类分别对环境、人类健康与动物福利的影响关系。针对二元作答选项（是/否）的题目，本研究构建**二元逻辑回归模型（binary logistic regression models）**，结果以阳性作答的**优势比（odds ratios）**及对应**95%置信区间（confidence intervals）**呈现；而针对有序作答变量（极负面、负面、中性、正面、极正面）的题目，则采用**比例优势回归模型（proportional odds regression models）**进行分析，结果以更倾向于正面作答的优势比呈现。在构建回归模型前，先对有序尺度变量进行归类合并：将两个最不倾向的选项合并为负面作答，两个最倾向的选项合并为正面作答，由此将每个有序变量划分为负面、中性、正面三类。所有题目均设置了“我不知道”的作答选项。因此，本研究同时分析了人口统计学因素对各题目相关认知情况（或认知缺失情况）的影响。本研究同时呈现基于单变量回归模型的未校正结果，以及多变量回归模型得到的校正结果。所有结论均基于5%的显著性水平。所有统计分析均采用版本4.3.1的R统计计算软件完成，其中有序回归模型的拟合借助R语言`ordinal`程序包实现。