Observational data

As explained in Causal Inference for Data Science

Observational data are measurements collected without assigning or controlling the conditions under which observations occur.

Observational data are collected without assigning or controlling the conditions under which observations occur. Unlike experimental data from an A/B test or RCT, these data record what happened in the world. They support description and prediction, but causal questions require extra care because treatment choices may be linked to other factors. section 1.1.1 section 1.4

Common forms include:

• Customer databases linking marketing efforts to customer actions. section 1.4
• Sales records used to examine changes after company decisions. section 1.4
• Cross-country data relating COVID-19 policies to virus spread. section 1.4

Kidney-stone treatments

A hospital compared two treatments using records from 700 patients rather than an RCT. Overall recovery favored treatment B, 81% versus 78%. After separating patients by stone size, treatment A performed better for both small and large stones. This Simpson’s paradox shows why a simple comparison can mislead when treatment groups differ in relevant characteristics. section 2.1 section 2.9

Figure 1.4 Both arcade revenue and computer science doctorates increase when the population increases, so the population is a potential common cause between the two.

A causal diagram can expose a potential common cause: population growth may increase both arcade revenue and computer science doctorates, making their strong correlation non-causal. section 1.4

How Causal Inference for Data Science covers it

How deep it goes

Explains collection, examples, confounding, stratification, DAGs, adjustment, and positivity, but defers detailed ATE computation.

Examples

Uses customer, sales, COVID-19, arcade-correlation, kidney-stone, and historical medical-treatment datasets.

What it assumes

Assumes familiarity with variables and treatment outcomes, while teaching the distinction between observational and experimental data from the beginning.

What you can do after

Evaluate/understand causal claims by identifying confounders, questioning correlations, and deciding when causal inference or experiments are needed.

How it connects

Connects observational data with experimental data, causal inference, confounding, Simpson’s paradox, DAGs, ATE, RCTs, and A/B tests.

liveBookmanning.com

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

Also explained in

• 

  liveBookmanning.com
  Delivering Machine Learning Projects: From design to deploymentObservational data is data collected from naturally occurring conditions without deliberately assigning interventions.
• 

  liveBookmanning.com
  Causal AIObservational data are measurements collected without the investigator assigning the treatment or intervention being studied.

Compare how 3 books explain observational data →

Related concepts

• causal hierarchy theorempart of · The causal hierarchy theorem states that observational, interventional, and counterfactual information are generally distinct levels that cannot be inferred from lower-level information without assumptions corresponding to the higher level.
• experimentationcontrasts with · Experimentation is the deliberate testing of an idea, intervention, or alternative by observing outcomes to gather evidence or compare effects.
• causal inferenceused by · Causal inference is the process of determining or estimating how causes, exposures, or interventions affect outcomes, using data and assumptions about causal relationships.
• causal adjustmentused by · Causal adjustment estimates a causal effect by conditioning on variables that block noncausal pathways and averaging over their target-population distribution.

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.