Bayesian modeling for pro-environmental behavior data: sorting and selecting relevant variables.

Pro-environmental behaviors towards climate change can be measured and evaluated in different fields. Typically, surveys are the standard tool for extracting personal information regarding this phenomenon. However, statistical modeling for these surveys is not straightforward, as the response variable is often not explicit. Hence, we propose a set of methodological procedures to deal with pro-environmental behavior data. First, validity evidence through a factorial analysis. Second, indexes are created from factor scores, where one of the latent factors summarizes a target variable. Third, a Beta regression is used to model the index of interest. Fourth, the inferential process is performed from a Bayesian perspective, in which posterior probabilities are used to sort and select the relevant variables. Finally, suitable models are obtained, and conclusions can be drawn from them. As a motivation, we used data from two Chilean surveys to illustrate our methodology as well as interpret and discuss the results.