7 Step 5: Estimation
7.1 Objective
Produce the canonical CU/WSP bundle used for release, review, and downstream metric execution.
7.2 What counts as the main output
The canonical analytical output is cu_timeseries.
You may also produce:
pop_timeseriesfor supporting representation,- a historical-context layer,
- a status bundle built from the CU series,
- a package or consumer bridge built after QC.
Do not let a consumer-specific file shape become the definition of the run. The CU series comes first.
7.3 Choose estimation mode
- Direct sum — near-complete site coverage with stable inclusion rules
- Indicator expansion — partial coverage with defensible expansion factors
- Run reconstruction / model-based — aggregate counts, stock-share data, or model structure required
7.4 Do this exactly
- build
cu_timeseriesfrom the processed records, - apply start-year policy where defined,
- populate the metric-facing fields deliberately,
- compare against the prior release,
- record any material difference in
release_diff_summary.md.
7.5 Minimum cu_timeseries contract
At minimum, the table should make these fields available or intentionally explain why they are absent:
CU_IDCU_NameSpeciesYearSpnForAbd_WildSpnForTrend_Wild
Strongly recommended context fields:
SpnForAbd_TotalSpnForTrend_TotalRecruits_WildRecruits_TotalGroup- method or provenance tag
7.6 Bundle files to keep together
| File | Role |
|---|---|
cu_timeseries |
canonical CU time series |
wsp_metric_specs |
one row per CU describing metric settings and benchmarks |
wsp_cyclic_benchmarks |
required only for cyclic benchmark cases |
cu_metadata |
supporting CU context, method, provenance, and interpretation |
published_integrated_statuses |
optional comparison input |
| sidecars | run, decision, QC, and metadata notes |
7.7 Alignment checks before release
- uniqueness by
(CU_ID, Year), - all CUs in
cu_timeseriesappear inwsp_metric_specs, - start-year trimming is documented,
- wild vs total semantics are consistent with the metric plan,
- cyclic benchmark files are present when required,
- prior-release diffs are explained in plain language.
7.8 If the estimation path is model-based
Add these fields or sidecars to the release package:
- point estimate,
- interval bounds,
cvorse,- method tag,
- run ID or control-file reference,
- short note on post-fit corrections.
That is enough to make the output reviewable without turning the cookbook into a full assessment manual.
7.9 Downstream metric execution
Preferred tooling:
https://github.com/Pacific-salmon-assess/WSP-Metrics-Pkghttps://github.com/SOLV-Code/WSP-Rapid-Status-WorkedExamples
Record the repo URL and commit SHA used for any downstream metric or rapid-status run. “Latest GitHub” is not a method.
7.10 Required outputs from Step 5
cu_timeseries.csv(or the species file mapped to that role)- optional
pop_timeseries.csv release_diff_summary.mdwsp_metric_specs.csv- optional
wsp_cyclic_benchmarks.csv