Subpopulation

As explained in Causal Inference for Data Science

A subpopulation is a subset of a broader population defined by shared characteristics or assignment conditions.

A subpopulation is a subset of a broader population selected because its members share characteristics or assignment conditions. In causal analysis, restricting attention to such a subset can make treatment and control cases more comparable, but changes the population to which the result applies. section 5.3.1 section 5.2

Selecting comparable patients

Suppose the control group contains only older patients while the treatment group contains both young and old patients. An analysis restricted to older patients may proceed, but its findings apply only to older people and cannot be extrapolated to younger people. When several confounders identify viable subsets, use their intersection as the subpopulation for further study. section 5.3.1

Propensity-score overlap provides another way to identify a subpopulation. Patients whose scores fall within a common-support interval have potential matches in the opposite group; patients outside it do not. Restricting analysis to the overlapping group may support an estimate for that group, but does not justify extrapolation to dropped patients. section 5.3.3

Figure 5.11 Partial overlap in a distribution’s supports. We can calculate the ATE only for patients within the common support region.

Use the common-support figure to see why partial overlap limits the ATE to matched patients; when overlap consists of treated patients, the estimand becomes the ATT. section 5.3.3

How Causal Inference for Data Science covers it

How deep it goes

Explains selection through confounder intersections and propensity-score overlap, while stopping short of prescribing a universal inclusion rule.

Examples

Uses older patients, confounder intersections, propensity-score intervals, and treated patients as concrete subpopulation examples.

What it assumes

Assumes treatment and control groups, confounders, propensity scores, and causal estimands such as ATE and ATT.

What you can do after

Evaluate whether a restricted causal estimate still matches the analysis objective and communicate its limited scope.

How it connects

Connects subpopulations to propensity scores, confounders, positivity, common support, ATE, and ATT.

liveBookmanning.com

Read Causal Inference for Data Science on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Causal AIA subpopulation is a defined subset of a broader population that shares specified characteristics or conditions.

Compare how 2 books explain subpopulation →

Related concepts

• populationpart of · A population is a defined group or complete set of people, units, observations, or candidate solutions considered together in a particular context.
• conditional probabilityused by · Conditional probability is the probability of an event or variable value given that specified information, an event, or another variable’s value is known.
• Simpson's paradoxcontrasts with · Simpson's paradox is the reversal or substantial alteration of an association between variables when data are aggregated rather than examined within relevant subgroups.
• set intersectionproduced by · Set intersection is the set of elements that belong to every one of two or more sets.
• extrapolationcontrasts with · Extrapolation is the estimation of values, outcomes, or conditions beyond the range represented by observed data by extending an inferred pattern or relationship.

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