Estimand

Option	Description
ATT	Average Treatment effect on the Treated: The mean effect for the units that actually received treatment.
ATE	Average Treatment Effect: The mean effect if the entire population were treated versus untreated.
LATE	Local Average Treatment Effect: The mean effect for units whose treatment status is affected by an instrument or cutoff.
CATE	Conditional ATE: The effect conditional on specific covariate values or subgroups.
Mediation effects	Decomposes the total effect into direct and indirect (mediated) components.
Oriented link	Identifies the existence and directions of associations (edges) in a causal graph without quantifying effect sizes.
Risk ratios	The ratio of outcome probabilities between treated and control groups (and derivatives).
Maps & generalisations	A model projection is the main method target, be it in space or time, rather than a causal effect.
Others	Any specialized causal or association measure not listed above.

Definition

The specific target quantity or causal parameter that the method seeks to estimate when precisely defined or more general quantities otherwise.

Explanation

Identifying the estimand clarifies what aspect of attribution or association the method recovers and for which subpopulation (e.g., all units, treated units only). This ensures alignment between your scientific question and the statistical target of inference.

Tools/rationale for helping assessment

Sketch a simple DAG of your system to identify which causal parameter answers your question. For more materials on causal effects definitions, see Igelström et al. (2022) Defining causal effects section, and Lundberg et al. (2021).

Example

• A landscape ecologist studying selective logging chooses ATT to estimate the average canopy-loss effect on logged forest pixels only and uses risk-ratios for bird occupancy comparisons.
• My DAG shows I care only about treated (disturbed) pixels’ biodiversity loss, so I pick ATT; I also want a spatial map of effects, so I note Maps & generalisations alongside.