W3 study map¶
The labels R76–R95 are research stage identifiers, not package versions, released modules or calibrated models.
| Series | Scientific question | Begin with |
|---|---|---|
| R76–R81 · Inference validity | Does leverage, covariance choice, time dependence, heavy tails or contamination undermine whole-function inclusion? | R76, R78, R80, R81 |
| R82–R83 · Observed gaze | Can a retained gaze fraction mask inadequate independent-unit support, and how do assumed MNAR mechanisms change sensitivity? | R82, R83 |
| R84–R88 · Validation of missing outcomes | When does genuinely new ascertainment help, and how can selective secondary return bias an audit? | R84, R85, R86, R88 |
| R89–R92 · Bounds and randomized designs | What can observable information and controlled audit assignment bound, without assuming guaranteed response? | R89, R91, R92 |
| R93–R95 · Exact identification | What remains unknowable under high response, and exactly which external guarantees or response-shape assumptions narrow the latent set? | R93, R94, R95 |
The arrows show the succession of questions; no stage in this diagram is a promotion of a failed procedure to the stable package.
Locate an original artifact¶
Evidence is indexed in the research evidence dashboard. Individual original-source figure ledgers and study-specific original/independent records retain the exact datasets, denominators, simulation laws, failure counts and CI references. Start with R93, R94 or R95.
Important boundary¶
No stage here claims successful general nominal-90% functional inference or observed-data identification of real-gaze MNAR. Current release status and issues #289 / #296 take precedence over chronology.