Results and assessment#
Use result objects to read estimates, uncertainty, influence curves, diagnostics, and provenance. Reports provide stable summaries and dataframe output for downstream analysis.
Define operations shared by fitted causal-result families. |
|
Store a fitted point-treatment TMLE result. |
|
Store estimates and fitted artifacts for longitudinal regimens. |
|
The result of a targeting step. |
|
One validation fold's contribution to a cross-validated targeting step. |
|
A point estimate with everything needed to do inference on it. |
|
Describe the structured identity behind a parameter alias. |
|
Describe whether and how a result supports one assessment operation. |
|
Collect validation, diagnostics, and sensitivity in one post-fit battery. |
|
Status returned by a diagnostic or validation operation. |
|
Collect statuses from a combined diagnostic or sensitivity run. |
|
Collect results from the inexpensive validation battery. |
|
Describe which post-fit actions a saved result can reproduce. |
|
One candidate exposure's target estimate and declared adjustment set. |
|
Candidate-exposure estimates with multiplicity-adjusted null tests. |
|
Estimate one causal importance parameter per candidate exposure. |
|
Load a complete result from a trusted joblib artifact. |
CausalResult is the protocol every fitted scalar result satisfies. The concrete classes are
TMLEResult for a point-treatment fit and LongitudinalResult for a sequential one; estimate()
returns one of those two.
TMLEResult.fluctuations returns one Fluctuation for each target group. A fold-targeted
cross-validated fit fills that record’s folds field with one FoldFluctuation for each
validation fold.