Bayesian project proposal template
Project title
State a concise title that names the scientific problem rather than only the model class.
Scientific question
- What is the substantive question?
- What is the target estimand?
- Why is a Bayesian approach appropriate?
Data and measurement context
- What are the observational units?
- How were the data collected?
- What dependence, grouping, missingness, or measurement issues are expected?
Candidate model family
Write the generative structure:
\[ p(y,\theta)=p(y \mid \theta)\pi(\theta). \]
If hierarchical or latent structure is involved, write the factorization in expanded form.
Prior strategy
- Which priors will be used?
- What scale information justifies them?
- What prior predictive checks will be run?
Computational plan
- Which inference method will be used?
- Why is it appropriate for the model geometry?
- Which diagnostics will be used?
Model ladder
List:
- the baseline model;
- the reference model;
- at least one sensitivity or robustness alternative.
Evaluation plan
- How will predictive performance be assessed?
- How will sensitivity analysis be conducted?
- What constitutes success for the project?
Risks
List the main expected failure modes:
- identifiability;
- poor mixing or divergences;
- prior sensitivity;
- weak signal;
- data quality problems.
Deliverables
- proposal;
- reproducible code;
- diagnostics;
- final report;
- appendix of sensitivity analyses.