The information collected covered organic practices adopted, societal and demographics of operators; characteristics of farms; farmers’ beliefs, attitudes, and perceptions related to organic farming practices, and so forth. A large sample of farmers from a population of total 4818farms in Georgia were randomly selected and interviewed. The instruments used in the interview were designed by faculty members of Fort Valley State University and the field work of the survey was administered by the Burruss Institute of Kennesaw State University. Farmers who respond to the screening question, “Do you produce fruits and vegetables?” with a “yes” were retained in the sample.About 2404 farmers were contacted, with 456 farmers going through the survey, which gave a return rate of 18.9%.
The instruments encompass abroad spectrum of questions pertinent to production practices, social demographics,individual attitude, beliefs, perceptions, as well as the characteristics of farms. The organic production in the survey may take the form of the USDA certified, certification exempt, or transitioning farms. The interview were conducted by trained personnel following the well-established procedures, which insures the veracity of data collected. However, it is also in evidence that some self-selection biases occurred due to the fact that the higher level of education were associated with organic producers and they were inclined to finish the survey retained in the sample, hydroponic dutch buckets which make the organic operations in the sample high than the overall percentage in Georgia farms in 2012 Census of Agriculture.The defect may constrain the effort to reach a general extrapolation beyond the survey data. In Table 1, the variables covered in the survey and the corresponding preliminary statistics were reported to provide a profile of small farmers in the Southern region of states.
We approached farmers’ choice of organic farming and potential factors of influence with the help of the logit regression model. After comparative study on the logit, the probit, and the linear probability models, being alike in ways in analyzing categorical data , we first exclude the linear probability model for its bias and inefficiency. The logit and probit models are arguably equivalent, only many investigators prefer the former for easy interpretation of parameters. For the sake of comparability, we used the logit model hereof. Since rich documents related to the logit model are readily available in the literature, the authors just present a model brief in the context of this investigation, rather than a thoroughgoing model discussion in the coming section. As usual, the most challenging part of modelling is associated with model selection among many alternatives. In this study, we adopted the approach of Purposeful Selection of Variables , which usually retains important confounding variables and result potentially in a slightly richer model. We beganour model fitting by a univariate analysis of all variable relevant. Any variable with a significant univariate test at 0.25 level was selected as a candidate for the multivariate analysis. In an iterative process, covariates are removed from the model if they are non-significant and not a confounder.
Significance is evaluated at the 0.1 level and confounding as a change in any remaining parameter estimate greater than 15% as compared to the full model. A change in a parameter estimate above the 15% indicates that the excluded variable was important in the sense of providing a needed adjustment for one or more of the variables remaining in the model. At the end of this iterative process of deleting, refitting,and verifying, the model contains significant covariates and confounders. Then,we took into account of any variable not selected for the original multivariate model and added them back one at a time, with all significant covariates and confounders retained earlier. In such a way, other variables which, by themselves,were not significantly related to the outcome but became an important contributor in the presence of other variables will be included in the final model.