Abstract:
|
Compared to the frequentist method, the Bayesian approach offers many advantages, as it is more intuitive, directly addresses the question of interest, properly accounts for uncertainty, allows the incorporation of utility for decision-making with multiple outcomes (e.g., efficacy and toxicity), etc. Bayes theorem is widely applied in assessing diagnostic accuracy by computing positive and negative predictive values. Under the Bayesian approach, prior information is easily incorporated for assessing treatment effect. A big challenge is how to communicate the Bayesian thinking to researchers who may not be well versed in statistics. In addition, there are relatively few tools for learning Bayesian update and gauging the impact of the prior distribution. We have developed a few Shiny applications that integrate R, HTML, and CSS for developing teaching tools. Shiny applications will be shown to illustrate Bayesian update via interactive graphs and videos in computing beta-binomial distribution, normal-normal distribution, normal-inverse gamma distribution, diagnostic test, ROC curve analysis, etc. The impact of the prior distribution can be easily displayed for the sensitivity analysis. Participants are encouraged to share their experiences and successful examples in teaching Bayesian methods to both statisticians and nonstatisticians.
|