Backdoor criterion

As explained in Causal AI

The backdoor criterion is a graphical condition requiring a covariate set to block every backdoor path from treatment to outcome while excluding descendants of the treatment.

When estimating how engagement changes in-game purchases, an observed association can mix the direct effect of engagement with information carried by a common cause. The backdoor criterion identifies which covariates remove that confounding so the intervention-level effect can be estimated. section 10.3.1

A covariate set satisfies the backdoor criterion when it d-separates every backdoor path from treatment X to outcome Y and contains no descendants of X. Adjusting for such a set blocks noncausal statistical signal while preserving the causal path of interest. section 10.3.1

Use the criterion to guide adjustment decisions:

• Block paths entering treatment through a common cause, such as E←G→I, by adjusting for G. section 10.3.1
• Choose any valid adjustment set; adjusting for every available covariate is neither necessary nor safe. section 11.4.1
• Avoid mediators, which block part of the causal effect, and colliders, which introduce collider bias. section 11.4.1
• After identification, estimate the effect with an appropriate model; linear regression remains subject to ordinary modeling considerations such as linearity. section 11.4.1

Gaming engagement and purchases

Suppose E is side-quest engagement and I is in-game purchases, while G is a common cause. Adjusting for Guild Membership, Time Spent Playing, and Player Skill Level blocks the relevant backdoor paths; Prior Experience is unnecessary because the other variables suffice, even though adding it would remain valid if observed. section 10.3.1 section 11.4.1

Figure 10.8 E←G→I is a backdoor path where G is a “confounder” that is a common cause of E and I. We are interested in the statistical signal flowing along the causal path from E to I, but that signal is “confounded” by the noncausal noise from additional statistical information through G on the backdoor path E←G→I.

Inspect the DAG showing how a common cause creates the backdoor path E←G→I and confounds the causal signal from E to I. section 10.3.1

How Causal AI covers it

How deep it goes

Defines path-blocking conditions, derives adjustment logic, and connects identification to regression while stopping short of estimator implementation details.

Examples

Uses an online gaming DAG, the path E←G→I, and alternative adjustment sets involving gameplay, guild membership, and player skill.

What it assumes

Assumes causal graphs and d-separation as working tools while developing the distinction between causal paths, backdoor paths, mediators, and colliders.

What you can do after

Apply it via DoWhy or a custom estimator to choose defensible adjustment variables before estimating a causal effect.

How it connects

Connects causal inference with d-separation, confounding, average treatment effects, mediation, collider bias, and linear regression.

liveBookmanning.com

Read Causal AI on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Causal Inference for Data ScienceThe back-door criterion is a graphical criterion for determining whether conditioning on a set of variables identifies a causal effect in a directed acyclic graph.

Compare how 2 books explain backdoor criterion →

Related concepts

• adjustment setused by · An adjustment set is a set of variables conditioned on to block noncausal paths between a treatment and outcome while preserving the causal paths.
• causal adjustmentconfigures · Causal adjustment estimates a causal effect by conditioning on variables that block noncausal pathways and averaging over their target-population distribution.
• backdoor pathuses · A backdoor path is a path between a putative cause and an outcome that begins with an arrow into the cause and can create a noncausal association.
• descendantuses · A descendant is a node or element reachable from another by following one or more downward links in a hierarchy or directed structure.
• d-separationhas part · D-separation is a graphical criterion for determining whether a conditioning set blocks all relevant paths between variables in a graph, implying conditional independence under the graph’s assumptions.
• directed acyclic graphused by · A directed acyclic graph is a graph with directed edges and no directed cycles.

This explainer was generated by Manning's knowledge graph from the text of Causal AI. The section links point to the parts of the book it draws on.