euromod_linking.methods.lma_labour_alignment.targets
====================================================

.. py:module:: euromod_linking.methods.lma_labour_alignment.targets

.. autoapi-nested-parse::

   Target construction: shock-table rows -> per-cell weighted targets.

   Cells are built from whatever dimensions appear in the shock groups: region x
   education, age x gender, or any other combination of input variables works
   identically. The region dimension is collapsed to the level the dataset
   actually supports (drgn2 -> drgn1 -> country) by truncating shock region codes
   to the data's code length and averaging growth-type shock values within the
   collapsed cell — an equal-weight average, since the microdata by definition
   cannot weight a level it does not resolve.



Attributes
----------

.. autoapisummary::

   euromod_linking.methods.lma_labour_alignment.targets.ALL_METRICS
   euromod_linking.methods.lma_labour_alignment.targets.LEVEL_CURRENT
   euromod_linking.methods.lma_labour_alignment.targets.MARGIN_OF
   euromod_linking.methods.lma_labour_alignment.targets.POOL_WARN_PCT
   euromod_linking.methods.lma_labour_alignment.targets.RATE_METRICS
   euromod_linking.methods.lma_labour_alignment.targets.RATE_NUMERATOR


Functions
---------

.. autoapisummary::

   euromod_linking.methods.lma_labour_alignment.targets.build_targets
   euromod_linking.methods.lma_labour_alignment.targets.cell_report
   euromod_linking.methods.lma_labour_alignment.targets.feasibility_warnings
   euromod_linking.methods.lma_labour_alignment.targets.totals


Module Contents
---------------

.. py:function:: build_targets(df: pandas.DataFrame, shocks_period: pandas.DataFrame, used: dict[str, list[str]], eligible: pandas.Series) -> tuple[pandas.DataFrame, list[str]]

   Per-cell current weighted counts + targets for employment/unemployment.

   The macro model's numbers are anchored on the *survey's own* baseline:
   a growth-rate shock becomes ``target = current_weighted_count * (1+rate)``,
   so only the macro model's projected *change* is imposed — any level
   disagreement between the macro model's employment stock and the weighted
   survey count (different definitions, reference periods) is deliberately
   not "corrected", which would move thousands of people for purely
   statistical reasons.

   Cells without a shock for a metric keep target = current: no shock, no
   artificial churn. Shock cells matching no data cell are reported as
   warnings (the macro model may cover regions the survey lacks).


.. py:function:: cell_report(cell_targets: pandas.DataFrame) -> list[dict]

   Per-cell baseline, resolved target, implied change and feasibility.

   The baseline is the number the alignment will actually move people away
   from, so a caller can check its own arithmetic against it *before* running:
   a growth rate derived from some other population count shows up here as a
   change that does not match what was asked for.

   ``change`` is reported in people, in percent, and in percentage points of
   the participation and inactivity rates — the last of these is usually the
   unit the request was phrased in ("inactivity down 3.5pp"), even when the
   shock itself was expressed as a level.

   Rates use the cell's working-age population as denominator; it and the
   shielded ``other`` state are invariant under the alignment (only the
   inactive <-> active margin moves), so the rates are directly comparable
   before and after. ``inactivity_rate`` is the conventional complement of the
   participation rate — everyone of working age who is neither employed nor
   unemployed, students and pensioners included — which is *wider* than the
   ``inactive_recruitable`` pool the methodology is able to move.


.. py:function:: feasibility_warnings(report: list[dict]) -> list[str]

   Warnings for cells whose shock consumes an implausible share of the pool.

   Only the ``inactive`` state can be moved, so a target derived from a broader
   notion of inactivity (one counting students, pensioners or everyone without
   employment income) asks for people who are not there to be moved.


.. py:function:: totals(cell_targets: pandas.DataFrame, accuracy: list[dict] | None = None) -> dict

   Scenario-wide sums over the shocked cells (unshocked cells contribute 0).


.. py:data:: ALL_METRICS

.. py:data:: LEVEL_CURRENT

.. py:data:: MARGIN_OF

.. py:data:: POOL_WARN_PCT
   :value: 25.0


.. py:data:: RATE_METRICS

.. py:data:: RATE_NUMERATOR