The frequentist statistical models that social work researchers typically use are implemented assuming that prior research does not inform our expectations about statistical trends in new studies. That is inconsistent with our ontological and epistemological beliefs of research. Consequently, many of our studies are underpowered for statistical testing: the estimated parameters are less precise, the null hypotheses cannot be rejected, and the replication probabilities are unknowable. Bayesian methods fundamentally address these deficiencies, by incorporating prior knowledge into current studies through probability distributions conditioning on observed data. Despite these advantages, there has been a minimal number of studies drawing on Bayesian methods to answer research questions in social work research. Moreover, the critical value-added of Bayesian methods to research, notably emphasis on evidence rather than error rates and accurate distribution of parameter estimates, has not been recognized enough. Thus, there remains a great need for a clear and concise overview of applying Bayesian methods to social work research.
Objectives
Upon completion of the workshop, participants will be able to:
1. recognize how the Bayesian approach overcomes deficiencies in the frequentist approach and understand the downside of Bayesian methods;
2. demonstrate understanding of the Bayes' theorem and evaluate its relevance to Bayesian inference in social work research;
3. use Stata to perform Bayesian regression analysis, including specifying priors, examining convergence of the model, and visualizing and interpreting posterior distributions;
4. evaluate the fit of the model and the implications of the resulting posterior distribution; and
5. demonstrate an understanding of the minimum reporting guidelines of Bayesian methods in social work research.
Content
Based on the pedagogical techniques of learning-science-by-doing-science and problem-based learning, this workshop aims to address the following content:
1. summary of advantages of Bayesian analysis over frequentist analysis and their similarities and differences;
2. introduction to Bayesian inference including Bayes' theorem, history, assumptions, and application;
3. a concise and systematic set of procedures to establish full probability Bayesian models;
4. application of Bayesian linear regression in social work research using a problem example based on real data;
5. a step-by-step demonstration of evaluation, interpretation, and visualization of model fit and posterior distribution using Stata; and
6. how to draw inferences and conclusions.
This workshop will encourage questions and comments throughout, and the pedagogical approach will exploit figures and illustrations to facilitate interpretation. All participants will receive a paper and/or electronic copy of a handout detailing the workshop content including slides, syntax, best practices, and a list of relevant resources.
Implications
Overall, this focus on Bayesian methods can be invaluable to social work researchers, faculty, students, and professionals who wish to refine their statistical analyses to derive richer inferences and conclusions. Featuring real data analyzed with a commonly used statistical program, this workshop simplifies esoteric statistical models making them more applicable to social work research. By becoming familiar with best practices in applying Bayesian methods, attendees can gain the expertise necessary to use a powerful set of tools that can help enhance data analysis and advance research in social work.