Averagers

This module provides implementations of unit averaging strategies.

The package implements several specific averaging schemes (optimal, mean group, individual-specific). It also provides the possibility of defining custom averaging scheme by inheriting from BaseUnitAverager and implementing appropriate weight schemes.

Available averagers:

Class	Description
BaseUnitAverager	Abstract base class. Subclass and implement _compute_weights() for custom strategies.
IndividualUnitAverager	Assigns all weight to the target unit (weights = [0, ..., 1, ..., 0]).
MeanGroupUnitAverager	Equal weights for all units (weights = [1/N, ..., 1/N]).
OptimalUnitAverager	MSE-optimal weights (agnostic and with prior restrictions).

All averagers share a common interface:

1. Fitting: Call fit(target_id) to compute weights and compute the unit averaging estimator for the focus function supplied to the class constructor.

2. Averaging: Use average() to recompute the averaging estimator

See also

• Tutorials for worked examples

• Focus Functions for information on focus functions.

• About Unit Averaging for theoretical background on the implemented schemes.