Designing a learning approach to community-based decision-making: Using analytical statistics and decision trees to optimize a data structure

This paper describes an effort to predict alumni success as measured by employer satisfaction in a veterinary medical education environment. Due to the complexity of the data, this paper sought to augment regression analysis with decision tree analysis to see if the combination of both approaches could result in new insights. Potential predictors of employer satisfaction were characterized as either technical or non-technical skills, and assigned to smaller conceptual item groupings within one of those areas. The internal consistency of each item grouping (concept) was then evaluated using Cronbach’s alpha. In the first approach, regression analysis was used to determine how well each concept predicted overall employer satisfaction with graduates. In the second approach, decision-tree analysis was used to discover the least uncertain hierarchy of this data, with overall employer satisfaction as the outcome. The process and the results from both approaches were examined and discussed along with recommended next steps.