Link to current text
MMP.7:3 - Forces
| Choice | What changes in the inference |
| Target population and included cases | A result about reported cases may require a selection model before it describes the target population. |
| Retain or remove unobserved quantities | Keeping them can clarify construction; summing or integrating them out can simplify computation. |
| Separate and shared influences | Conditional independence can hold while observations remain dependent after a shared influence is removed. |
| Rich observation model and obtainable information | Extra parameters can represent real effects while leaving the desired answer less identifiable. |
| Probabilistic answer and sufficient conditional answer | A likelihood can be useful before choosing an estimator or a prior. A bound may already settle the action. |