1/109
Comprehensive vocabulary flashcards covering causal models, sample representation, external validity, sample selection bias, mediation, moderation, confounding, and collider bias.
Causal Research Question
A research inquiry focused on determining whether a change in an independent variable (X) causes a change in a dependent variable (Y).
Population
The entire group of individuals, units, or cases (such as all humans, Americans, or American high schoolers) about which social scientists seek to draw conclusions.
Sample
A selected subset of a population used in conjunction with statistical theory to infer characteristics of the larger population when full data is unavailable.
Statistical Theory
The mathematical framework used by social scientists to draw valid inferences about a larger population from a representative sample.
Sample Representation
The condition where a study sample accurately mirrors the characteristics of the target population; a requirement often violated in social science research.
External Validity
The degree to which a research finding can be generalized beyond the specific study sample to a broader population.
Oversampling
The selection of a higher proportion of a specific demographic or subgroup in a sample (e.g., 40/60 Male/Female) than exists in the actual population (e.g., 50/50 Male/Female).
Downweighting
A statistical adjustment procedure in which oversampled respondents are assigned lower mathematical weights to produce an unbiased population estimate.
Sample Selection Bias
Inaccurate results produced because selection into the study sample depends on the value of the outcome, or on factors related to both the independent variable and the outcome.
Diagnostic Question for Sample Selection Bias
The critical evaluative question: "Could a case’s value on the outcome have affected whether it ended up in my sample?"
LinkedIn Sampling Example
A study design flaw where sampling LinkedIn users to examine education’s impact on income creates sample selection bias because higher earners disproportionately use LinkedIn.
External Validity vs. Sample Selection Bias
External validity asks whether sample results transfer to a larger population, whereas sample selection bias means the results can be incorrect even for the specific sample studied.
Accuracy vs. Representativeness in External Validity
A principle noting that a study lacking external validity can still produce accurate, internally valid results for the sample, even though they fail to represent the broader population.
Causal Model
A conceptual diagram describing how an independent variable relates to a dependent variable, as well as the paths through which other variables influence or are influenced by them.
Connecting Paths
Directed arrows in a causal model that represent hypothesized causal impacts and relationships among variables.
Absence of Connecting Paths
The deliberate omission of lines or arrows in a causal model that explicitly describes ways an independent variable does not relate to a dependent variable.
Blueprint Function of Causal Models
The foundational role of causal models in guiding researchers on which measures to include or exclude in statistical models to estimate unbiased relationships.
Unbiased Relationship Estimation
The statistical outcome of correctly including true confounders and excluding colliders to accurately test causal hypotheses.
Independent Variable (D or X)
The presumed causal variable in an empirical study whose effect on a dependent outcome is under investigation.
Dependent Variable (Y)
The outcome variable hypothesized to be influenced or caused by changes in the independent variable.
Program Evaluation Model
A structured causal model mapping programmatic service activities through intermediate victim and investigator mechanisms to legal outcomes.
Victim Service Provider
The primary intervention agency in the program evaluation example delivering assistance, crisis counseling, and advocacy to crime victims.
Victim's Compensation Filing
A service provider activity that assists crime victims with filing claims for financial compensation.
CJ Support/Advocacy
Criminal justice system support and personal advocacy services provided to victims navigating legal procedures.
Crisis Counseling
Immediate psychological and emotional assistance offered to victims following a traumatic incident.
Counseling/Support
Ongoing therapeutic and emotional support provided by victim service agencies to foster recovery.
Information and Referral
The provision of resource links and guidance directing crime victims to relevant community services.
Personal Advocacy
Direct individualized assistance offered by advocates to secure victim rights and accommodations.
Other Case Management Support
Administrative and logistical aid delivered by providers to manage victim cases comprehensively.
On-Scene Support
Immediate, in-person crisis aid delivered by service providers directly at the scene of a crime.
Victim Engagement
An intermediate mechanism in the program evaluation model encompassing trauma mitigation, trust building, and meeting basic needs.
Trauma Mitigation
A sub-process of victim engagement focused on reducing the psychological impact of victimization.
Trust Building
A sub-process of victim engagement that establishes rapport and cooperation between the victim and authorities.
Meeting Basic Needs
Fulfilling urgent, foundational victim requirements (e.g., food, shelter, safety) to facilitate engagement.
Investigative Capacity
An intermediate law enforcement mechanism enhanced when victim advocates relieve detectives of supportive tasks.
Freeing Up Investigators
The direct programmatic benefit of reallocating non-investigative tasks from detectives to victim service providers.
New Investigative Leads
Actionable case details and evidentiary clues generated through victim trust, engagement, and freed-up investigative hours.
Arrest & Adjudication
The primary final criminal justice outcome achieved following successful investigation and lead generation.
Direct Effect (Causal Models)
A causal relationship where a change in one variable produces a change in another variable with nothing operating in between.
Mediator
An intervening mechanism or variable that links a cause to an effect and explains how or why the relationship occurs.
Indirect Effect
A causal impact of an independent variable on an outcome that operates through one or more intermediate mediators.
Moderator
A variable that strengthens, diminishes, or changes the direction of a cause-and-effect relationship without explaining the underlying mechanism.
Confounding Variable
A common cause of both the independent variable and dependent variable that produces a spurious association if not controlled.
Collider
A variable that is a common effect of both the independent variable and dependent variable.
Observed Variable
A factor in a causal model that is directly measured and empirically captured in a dataset.
Unobserved Variable
A latent or unmeasured factor (such as an individual's true threat level) that influences other variables in a causal model.
Uni-directional Relationship
A one-way causal path where variable A causes changes in variable B, but B does not influence A.
Bi-directional Relationship
A reciprocal or feedback loop relationship where two variables mutually affect each other over time.
Time Component
The chronological requirement in causal modeling dictating that a cause must precede an effect in time.
Temporal Precedence
The essential principle that an explanatory variable must occur prior to the observed change in the outcome variable.
Direct Effect (Mediation Definition)
An unmediated effect of a cause on an outcome indicating that no intermediate variable operates between them.
Indirect Effect (Mediation Definition)
An effect of a cause on an outcome that is transmitted through one or more mediating factors.
Unmediated Pathway
A direct causal path connecting cause to effect without reliance on intermediary steps or mechanisms.
Imprisonment Direct Effect Example
The reduction of community crime caused directly by incapacitation, because the imprisoned person has no physical access to the community.
Imprisonment Indirect Effect Example
The reduction of community crime caused indirectly because an individual receives rehabilitation support services while in prison, leading to law-following upon release.
Total Effect
The overall impact of a cause on an outcome, calculated as the sum of all direct (unmediated) and indirect (mediated) effects.
Growth of AI Mediation Example
A conceptual model where the growth of AI impacts poorer grades indirectly through reduced student learning, as there is no direct physiological effect on learning.
Reduced Student Learning
The mediating variable in the AI model that explains why increased AI usage results in poorer student grades.
Poorer Student Grades
The terminal outcome variable in the educational mediation model caused by reduced learning.
Thornberry et al. (1994)
A longitudinal study testing interactional theory by examining whether delinquent peers, beliefs, and behavior operate via selection, socialization, or interaction.
Interactional Theory
A criminological theory asserting that delinquent peers, delinquent beliefs, and delinquent behavior mutually reinforce one another bidirectionally over time.
Delinquent Peers
Peer networks engaged in antisocial conduct, examined across ages 12, 14, and 16 in longitudinal delinquency models.
Delinquent Beliefs
Attitudes and rationalizations favorable toward rule-breaking behavior, tracked longitudinally across adolescent development.
Delinquent Behavior
Actual participation in delinquent and criminal activities measured repeatedly over time in tests of interactional theory.
Selection Model (Thornberry et al.)
A criminological hypothesis where prior delinquent behavior and beliefs drive an individual to select delinquent friends ("birds of a feather flock together").
"Birds of a Feather Flock Together"
An idiom characterizing the selection hypothesis, indicating that individuals seek out peers who share their preexisting antisocial beliefs and behaviors.
Stability Effects
The longitudinal persistence of a single construct over time, such as delinquent behavior at Age 12 directly predicting delinquent behavior at Age 14.
Lag Effects
Cross-lagged causal pathways across longitudinal waves where one variable at an earlier age influences a different variable at a later age.
Socialization Model (Thornberry et al.)
A criminological hypothesis stating that association with delinquent peers causes and reinforces subsequent delinquent beliefs and delinquent behavior.
Reinforcement in Socialization
The social learning mechanism by which delinquent peer groups encourage, normalize, and reward law-violating attitudes and actions.
Interactional Model (Thornberry et al.)
The empirical model demonstrating that delinquent peers, beliefs, and behaviors all reciprocally affect one another across adolescence.
Apel & Horney (2017)
A research study investigating the causal relationship between employment and criminal offending, including the underlying explanatory mechanisms.
Employment and Crime Research Question
The central inquiry of Apel & Horney (2017): "Does being employed decrease the likelihood of criminal offending, and, if so, why?"
Life Controls (Apel & Horney)
Situational and domestic control variables in Apel & Horney (2017), including community corrections status, living arrangements, and school enrollment.
In Community Corrections
A life control variable in Apel & Horney (2017) tracking whether an individual is currently supervised within a community-based correctional sanction.
On Probation or Parole
A formal legal status variable controlled for in Apel & Horney's (2017) employment and crime model.
Domestic Cohabitation Life Controls
Variables in Apel & Horney (2017) capturing whether a participant is living with a wife or living with a girlfriend.
Family Cohabitation Life Controls
Variables in Apel & Horney (2017) measuring whether a subject resides with other family members or living with children.
Enrolled in School
An educational attachment control variable included in Apel & Horney's (2017) framework predicting criminal offending.
Routine Activities (Apel & Horney)
An intervening mechanism in Apel & Horney (2017) reflecting structured daily patterns and time use that mediate the employment-crime relationship.
Individual Differences (Apel & Horney)
Person-specific baseline traits and propensities accounted for when estimating the causal impact of employment on crime.
Moderation
A causal structure where a third variable changes the strength or direction of the relationship between an independent and dependent variable.
Mediator vs. Moderator Distinction
The conceptual distinction stating that mediators explain how or why a cause produces an effect, while moderators change the relationship and do not explain it.
"Old Dog New Tricks" Idiom
A saying representing age as a moderator that alters the effectiveness of an educational or behavioral intervention.
AI Competence Class Example
An applied example where age moderates the impact of an AI class, leading to increased AI use among younger attendees but not older attendees.
Interaction Term
A statistical product variable created by multiplying a potential moderator by the independent variable to test for moderation in a regression model.
Dual Role of Gender
The causal scenario where gender acts simultaneously as a moderator of education’s effect on income and as a confounder of their association.
Co-production
The cooperative behavior wherein citizens assist public officials in the delivery and provision of public services.
Letki & Steen (2020)
A study examining the incentives and environmental factors that impact an individual's willingness to help public officials provide public services.
Willingness to Co-produce
The primary dependent outcome in Letki & Steen (2020) reflecting citizen readiness to participate in public service co-production.
Environmental Factors in Co-production
Contextual community variables modeled in Letki & Steen (2020), including community attachment, ethnic diversity, unemployment, and population density.
Attitudinal Factors in Co-production
Perceptual and institutional variables in Letki & Steen (2020), including trust in local authorities and perceived utility of action.
Incentive Structures in Co-production
Motivational drivers analyzed by Letki & Steen (2020), consisting of reputational rewards, monetary rewards, and social norms.
Confounder (Definition and Solution)
A common cause of both the independent variable (D) and dependent variable (Y); the solution is to control for it (X) to obtain an unbiased estimate.
Collider (Definition and Solution)
A common effect of both the independent variable (D) and dependent variable (Y); the solution is to avoid controlling for it (X) to prevent bias.
Conditioning on a Collider
The statistical error of controlling for or stratifying by a common effect (X), which induces a spurious correlation between independent causes (D and Y).
Workplace Grievance Collider Example
A collider scenario where perceived discrimination (D) and low commitment (Y) both cause grievance filing (X); controlling for filing distorts the true D→Y relationship.
Grievance Filer Asymmetry
The mechanism where filers who don't perceive discrimination are disproportionately low-commitment, while non-filers perceiving discrimination are high-commitment, biasing conditioned samples.
Collider Bias via Study Participation
Sample selection bias created when volunteering to participate in a study (X) is a common effect of employment (D) and criminal behavior (Y).
Employment and Crime Collider Reversal
A demonstration where volunteering bias reverses a true negative population relationship (70% employed, 20% criminal) into an artificial positive sample relationship (90% employed, 5% criminal).