Workflow articles

Twenty Python article companions cover the scope of the frozen gp3ml 0.3.0 vignette set. Use this page by research need, not by filename order.

End-to-end

Integrated research workflow

Move from task declaration and feature provenance through grouped evaluation, evidence capture, and reporting.

Generalization

Participant generalization

Build and validate models when assessment participants must be unseen during fitting.

Governance

Decision governance

Keep thresholds, asymmetric costs, calibration sources, and abstention explicit.

Validation

External validation reporting

Declare a second dataset, quantify transportability, and report it without contaminating development.

Pick a learning track

Track 1

Build the analysis correctly

1. Integrated research workflow
2. Analysis-plan governance
3. Assigned-condition discrimination
4. Recording-quality review

Track 2

Make the generalization claim explicit

1. Participant generalization
2. Stimulus generalization
3. Participant–stimulus generalization
4. Nested grouped resampling
5. Group-aware conformal prediction

Track 3

Govern decisions and validation

1. Decision governance
2. External validation reporting
3. Dataset shift and robustness
4. Contaminated-manifest leakage
5. Observed behavioral endpoint

Track 4

Harden reproducibility and portability

1. Reproducibility hardening
2. Portable research artifacts
3. Cross-package interoperability
4. Optional-engine portability
5. API stability contracts

Complete article map

Governance and planning

Article	What it helps you do
Analysis-plan governance	Declare, lock, validate, and audit a modelling plan before outcome-driven changes accumulate.
Governance standards profile	Map gp3ml evidence to a governance profile without claiming certification or external endorsement.
Decision governance	Make thresholds, asymmetric costs, calibration sources, and abstention explicit.
API stability contracts	Inspect stable versus experimental interfaces and audit a frozen public API.

Generalization and resampling

Article	What it helps you do
Participant generalization	Keep assessment participants unseen during fitting.
Stimulus generalization	Evaluate performance on stimuli not used during fitting.
Participant–stimulus generalization	Enforce both participant and stimulus independence.
Nested grouped resampling	Separate inner tuning from outer assessment while respecting grouping.
Group-aware conformal prediction	Add prediction sets or intervals while retaining the declared grouping unit.

Validation, shift, and robustness

Article	What it helps you do
External validation reporting	Declare an external dataset, evaluate transportability, and report it explicitly.
Dataset shift and robustness	Separate covariate and missingness shift from performance and calibration degradation.
Contaminated-manifest leakage	See how provenance and role checks detect predictors that should not enter the model.
Observed behavioral endpoint	Keep the endpoint tied to an explicitly observed, permissible behavioral quantity.

Reproducibility and interoperability

Article	What it helps you do
Reproducibility hardening	Capture environment differences and audit volatile artifacts.
Portable research artifacts	Package model and evidence outputs with safer persistence and provenance expectations.
Cross-package interoperability	Create typed handoffs between gp3mlpy and adjacent research tooling.
Optional-engine portability	Check engine capabilities and make optional-backend assumptions explicit.

Runnable and visual companions

Every article has a corresponding Python script under the repository's examples/ directory. CI executes the example suite so article-facing workflows remain coupled to the package implementation.

Several workflow families also produce auditable plot objects. The plot gallery shows generated examples for decision thresholds, dataset shift, engine portability, governance evidence, and other registered plot contracts.