Published: 13 January 2023| Version 1 | DOI: 10.17632/2vnt5pw33j.1
Sheila Garcia


Original Data Set of the research. A descriptive and quantitative research scheme, including a structured questionnaire adapted from previously-validated scales. Using data collected online from 281 rural producers in Brazil.


Steps to reproduce

Data collection took place between April 19 and July 3, 2021, using an Google Forms service. The target population consisted of rural producers, farmers or ranchers, individuals or legal entities, and hired rural administrators (decision maker). The lack of a definitive list of rural/livestock producers in Brazil was a barrier to a probabilistic sampling. A filter question verified whether the person willing to respond to the questionnaire fit one of the population profiles previously mentioned and accepted in the survey. The questionnaire was forwarded by the second author, directly and personally, via messaging applications and e-mails, to more than 1,000 potential respondents, a characteristic of selection by judgment. Each recipient was then asked to forward the invitation to their circle of acquaintances (the snowball technique). The link to the online questionnaire offered a brief description of the research and presented the researchers. 360 questionnaires were completed and, after an initial filtering process, 310 valid questionnaires remained. To assess possible biases arising from the data collection process, the Common Method Bias (CMB) was calculated. We used the Harman single-factor test, with exploratory factor analysis without rotation and with a single factor, whose extracted variance must be less than 50% (HAIR et al., 2019, p. 744). Statistical tests were performed to analyze the normality of the sample and its suitability for structural equation modeling using the PLS method (Mardia's test). Mardia's PK test was performed (Lisrel 8.8 software). The test based on a chi-square distribution (asymmetry and kurtosis) returned a p-value of less than 0.0001, indicating that the data did not meet the assumption of multivariate normality required by SEM based on covariance. Thus, SEM-PLS (based on correlation) was used, which does not require this assumption (HAIR et al., 2014b). We started from scales that had already been validated in other contexts. The measurements are given on a seven points Likert-type scale. The validation in Portuguese of the scales followed a series of steps (DEVELLIS, 2003): 1) submission of the adopted scale to the reverse translation technique; 2) theoretical or face validation by three marketing specialists, who helped to adjust the phrasing and evaluated the fit of the items in the constructs; 3) semantic validation with 40 respondents from the target population, in a face-to-face pre-test; and 4) statistical validation through composite confirmatory analysis associated with partial least squares structural equation modeling (PLS-SEM). With Structural Equation Modeling (SEM) through the SmartPLS 2.0 software, we evaluated, as a first step, the properties of the measurement models of each construct. In a second step, we evaluated the structural relationships of the Conceptual Model.


Universidade Estadual Paulista Julio de Mesquita Filho Faculdade de Ciencias Agrarias e Veterinarias Campus de Jaboticabal


Social Media, Consumer Attitude, Rural Marketing