Research questions, study design, and estimands

Define the target and establish what the observations identify

A research question needs a population, an independent unit, a measurement scale, and a quantity to estimate. Study design and observation determine which comparisons are supported. A regression formula alone does not specify those choices.

The running example asks for the mean measured feature level among eligible people when some outcomes are missing. It separates that population from the observed subset, then uses simple strata to show what identifying assumptions supply and what absent support leaves unknown.

What you’ll learn

• Specify the scientific target: name the population, unit, time, scale, contrast and weighting, and distinguish it from a procedure’s limit.
• Establish identification: express the target using the observed distribution under explicit assumptions, or construct compatible populations with different answers.

Completion means being able to write an estimand and identification specification for a research question, including an unsupported claim or missing piece of information. The page checkpoints exercise this through replication, target mismatch, standardization, and missingness; additional practice is optional.