Research Methods Program

Department of Biostatistics

Vanderbilt University School of Medicine

Nashville TN USA

- Bayesian fully sequential design and analysis plan for the 7-level COVID ordinal outcome scale for rapid learning and decision making
- Discussion about the statistical model for the outcomes and related Bayesian modeling discussion
- General rationale and methods for Bayesian clinical trial design and analysis and its overview
- Nathan James’
`R`

package for the Bayesian proportional odds model - Stan code from Ben Goodrich (Columbia University) for the PO model and Ben’s notes
- Stan code repository from Ben Goodrich
- Continuous learning from data: No multiplicities from computing and using Bayesian posterior probabilities as often as desired
- Discussion Board, which includes new updates regarding
`Stan`

code for the Bayesian PO model - Discussion Board for COVID-19 research methods resources in general
- R code example of a frequentist mixed effects ordinal PO model vs. a Bayesian random effects ordinal model for longitudinal data
- Papers about statistical modeling of longitudinal ordinal responses
- Biostatistics in Biomedical Research Tutorials - see sections below:

Topic |
Section |
---|---|

Ordinal outcomes in clinical trials | 3.6 and 5.12.5 |

Proportional odds model | 7.6 |

Power calculations tailored to the proportional odds model | 7.8.3 |

Bayesian approach for a single mean | 5.6.2 |

Bayesian logistic model | 6.10.3 |

Bayesian two-sample t-test | 5.9.3 and 5.10.5 |

Analysis plan for differential treatment effect (heteroeneity of treatment effect) and why subgroup analysis should be avoided | 13.6 |

Covariate adjustment in RCTs | Chapter 13 |

Overview of branches of statistics | 5.3 |

Problems with p-values | 5.4 |

- Information gain from using ordinal instead of binary outcomes
- Bayesian vs. frequentist statements about treatment efficacy
- Journey from frequentist to Bayesian statistics
- A litany of problems with p-values
- P-values and type I errors are not the probabilities we need
- Null hypothesis significance testing never worked