Research Methods and Causal Inference Vocabulary Flashcards

Vocabulary flashcards covering core concepts in research design, causal inference, experimental and observational methods, measurement operationalization, and sampling theory.

R Commander

A graphical user interface (GUI) for R that provides a point-and-click environment resembling Stata to facilitate statistical analysis without requiring extensive coding.

Intervening Variable (Causal Mechanism)

A variable ($Z$) positioned between an independent variable ($X$) and a dependent variable ($Y$) that explains the process or story of how $X$ influences $Y$.

Antecedent Variable

Another term for an independent variable ($X$) that precedes an intervening variable ($Z$) and a dependent variable ($Y$) in a theoretical causal chain.

Reverse Causation

A causal pitfall occurring when the assumed dependent variable ($Y$) actually causes the independent variable ($X$), or when $X$ and $Y$ co-influence each other simultaneously.

Spurious Relationship

An apparent relationship between two variables ($X$ and $Y$) that is non-causal and completely vanishes once an unmeasured or confounding variable ($Z$) is accounted for.

Confounding Variable

An unaccounted-for variable ($Z$) that influences both the independent variable ($X$) and the dependent variable ($Y$), creating a false appearance of a causal link between $X$ and $Y$.

Four Hurdles of Causality

A four-step framework for establishing cause-and-effect: 1) credible causal mechanism, 2) ruling out reverse causation, 3) establishing covariation between $X$ and $Y$, and 4) controlling for confounding variables ($Z$).

Field Experiment

An experimental study conducted in a real-world setting—such as a school, business, neighborhood, or country—rather than a laboratory.

Survey Experiment

An experiment embedded within a survey where respondents are randomly assigned different versions of questions or prompts, often using a split-ballot design.

Split-Ballot Design

A survey experiment technique where different respondents receive different survey versions according to a random probability assignment.

Experimental Arms

The distinct treatment and control groups or conditions structured within an experimental design.

Dose-Response Design

An experimental layout featuring multiple treatment arms that vary the dosage or degree of an independent variable to evaluate changing effects on an outcome.

Power Analysis

A statistical calculation used prior to conducting an experiment to determine the minimum number of subjects or units needed per group to detect an expected effect size.

Natural Experiment

A study design that takes advantage of an exogenous, naturally occurring, or unplanned event (such as a natural disaster or policy shock across borders) that assigns treatment independently of the outcome.

Quasi-Experiment

A research design where a treatment variable ($X$) is manipulated or experienced, but subjects cannot be randomly assigned by the researcher, requiring constructed comparison groups.

Monopsony

A market condition characterized by a single buyer—such as a sole employer in an isolated geographic area—who holds power over market terms for sellers or workers.

Random Assignment

The experimental process of randomly sorting subjects into treatment or control groups, ensuring high internal validity by balancing confounding variables.

Random Sampling

The selection process in observational research where every member of a population has an equal probability of selection, maximizing external validity and generalizability.

Ecological Fallacy

The logical error of making inferences about individual behavior or characteristics based solely on aggregate or group-level data.

Ecological Inference

The methodological process of using aggregate-level data to infer or estimate relationships at the individual level.

Regression to the Mean

The statistical phenomenon where an extreme value of a variable on an initial measurement is followed by a value closer to the average on subsequent measurements due to the balancing out of luck or random chance.

Synthetic Control Method

A quantitative comparative technique that uses linear programming to construct a weighted combination of untreated control units to mimic a treated unit's pretreatment characteristics and trajectory.

Prospective Study

An observational study that identifies a group or cohort and follows it forward over time to track outcomes following a specific event or intervention.

Retrospective Study

An observational study that starts with an observed outcome and looks backward in time to collect historical data on exposures or background characteristics.

Panel Data

Data collected by observing the exact same cross-sectional units (such as individuals, firms, or states) repeatedly over multiple consecutive time periods.

Operationalization

The process of systematically observing a theoretical concept in the real world and recording it as a measurable, quantitative variable.

Manifest Construct

A feature or variable that can be directly observed and recorded without needing indirect estimates, such as height, weight, or formal institutional counts.

Latent Construct

An abstract concept or underlying factor that cannot be directly observed—such as self-esteem, ideology, or social support—and must be measured via proxy or indicator variables.

Proxy Variable

An indirect, approximate quantitative measure used in place of an unobservable latent variable, such as using free and reduced lunch rates to approximate family income.

Indicator Variable

An observable measure used to signal or capture specific aspects of an abstract latent concept.

Measurement Scale (Index)

A composite variable constructed by combining multiple indicator questions or observations into a single summary score, such as the Consumer Price Index or GPA.

Measurement Validity

The extent to which an operationalized measuring instrument accurately captures the theoretical construct it was designed to measure.

Face Validity

A basic evaluation of whether a measurement instrument appears, on its surface, to be a reasonable measure of the intended construct.

Content Validity

The degree to which a measurement instrument covers all necessary dimensions and domains of the target theoretical concept.

Construct Validity

The extent to which an operational measure captures the targeted theoretical concept rather than measuring related or confounding sister variables.

Population (Universe)

The complete group of individuals, units, or occurrences that a researcher aims to investigate based on their theoretical model.

Sample

A subset of units or individuals selected from a broader population to be directly observed and analyzed in a study.

Statistical Inference

The process of using sample parameters to draw formal conclusions or estimates about unknown parameters of the broader population.

Simple Random Sample (SRS)

A fundamental sampling technique where every individual or element in a population has an equal probability of being selected.

Sampling Error

The expected variation or distance between a sample statistic and the true population parameter caused purely by luck of the draw.

Central Limit Theorem (CLT)

A core statistical principle stating that as sample size increases, the sampling distribution of the sample mean approaches a normal distribution regardless of the underlying population's shape.

Standard Normal Distribution

A symmetric, bell-shaped theoretical distribution centered at its mean, where mean, median, and mode are equal, and where observations follow the $68\%-95\%-99\%$ rule.

Sampling Distribution

The theoretical probability distribution of a statistic (such as a sample mean) generated by taking an infinite number of repeated samples of size $n$ from a population.

Standard Error

A statistical measure of the variability or spread of sample estimates across a sampling distribution, which decreases as sample size increases.

Sample Size Rule of Thumb Formula

The rule-of-thumb equation $n = \left(\frac{1}{E}\right)^2$ used to calculate the required sample size $n$ given a desired margin of error $E$.