PBSI 302: (Textbook)

Comprehensive practice flashcards covering longitudinal designs, cross-sectional correlations, multiple regression, pattern and parsimony, mediation, and validity evaluation.

Bivariate Correlation

An association that measures only $2$ variables and cannot prove causation on its own.

Multivariate Correlational Design

A research design that measures more than $2$ variables and examines their relationships without manipulating any variables.

Three Major Approaches to Multivariate Correlational Research

Longitudinal designs (addressing timing), multiple regression (ruling out measured third variables), and pattern and parsimony (gathering converging evidence).

Experiments vs. Correlational Research

Studies where researchers manipulate a variable and measure the outcome; random assignment usually provides stronger causal evidence than correlational research.

Covariance

A causal criterion requiring that the proposed cause and effect are associated and vary together.

Temporal Precedence

A causal criterion requiring that the proposed cause comes before the proposed effect in time.

Internal Validity

A causal criterion requiring that the association between two variables cannot be explained by a plausible third variable or alternative explanation.

Narcissism

A trait defined by feeling superior, deserving special treatment, and reacting strongly to criticism, which is distinct from healthy high self-esteem.

Parental Overvaluation (Overpraise)

A parenting practice where parents tell children they are more exceptional or special than other children, distinct from parental warmth and love.

Otway and Vignoles (2006)

A study demonstrating covariance between recalled frequent parental praise and adult narcissism ($r \approx .20$).

Temporal Precedence Issue in Otway and Vignoles (2006)

The limitation arising because adults rated current narcissism and recalled earlier parenting in a single session, meaning current beliefs might bias memories.

Third-Variable Problem in Otway and Vignoles (2006)

The possibility that external factors, such as gender or parents' own narcissistic tendencies, are related to both parental overpraise and child narcissism.

Longitudinal Design

A research design that measures the same variables in the same people at several points in time to evaluate changes and provide evidence for temporal precedence.

Brummelman et al. (2015) Sample and Waves

A longitudinal study of $565$ Dutch children and their mothers and fathers, contacted $4$ times, every $6$ months (about $2$ years).

Eight Variables in Brummelman et al. (2015)

The multivariate set of variables measured across time: child narcissism at Times 1–4 and parental overvaluation at Times 1–4.

Cross-Sectional Correlation

A correlation in a longitudinal design that tests whether two different variables measured at the same point in time are correlated.

Directionality Limit of Cross-Sectional Correlations

The inability of same-time measurements to establish temporal precedence, because either variable might have led to changes in the other.

Brummelman et al. (2015) Time 1 Cross-Sectional Correlation

The cross-sectional correlation between parental overvaluation and child narcissism at Time 1, which was $r = .007$.

Brummelman et al. (2015) Time 2 Cross-Sectional Correlation

The cross-sectional correlation between parental overvaluation and child narcissism at Time 2, which was $r = .070$.

Brummelman et al. (2015) Time 3 Cross-Sectional Correlation

The cross-sectional correlation between parental overvaluation and child narcissism at Time 3, which was $r = .138$.

Brummelman et al. (2015) Time 4 Cross-Sectional Correlation

The cross-sectional correlation between mothers' overvaluation and child narcissism at Time 4, which was $r = .099$.

Autocorrelation

In a longitudinal study, the correlation of the same variable measured at different points in time, indicating its stability over time.

Test-Retest Reliability and Autocorrelation

The measurement concept related to autocorrelation that assesses consistency over time when measuring the exact same construct across waves.

Cross-Lag Correlation

A longitudinal correlation between one variable at an earlier time and a different variable at a later time, helping establish directionality and temporal precedence.

Brummelman et al. (2015) Cross-Lag Directionality Findings

Results demonstrating that earlier parental overpraise significantly predicted later child narcissism.

Brummelman et al. (2015) Reverse Cross-Lag Path

Results demonstrating that earlier child narcissism did not significantly predict later parental overpraise.

Maternal vs. Paternal Overvaluation Results

Separate analyses of mothers' and fathers' overvaluation in Brummelman et al. (2015) showing similar causal path patterns for both parents.

Mutually Reinforcing Cycle

A cross-lag correlation pattern where both variables significantly predict each other over time in a reciprocal feedback loop.

Three Possible Cross-Lag Patterns

The three outcomes that can occur in cross-lag analyses: variable $A$ leads to variable $B$ only; variable $B$ leads to variable $A$ only; or both directions are significant.

Child Gender Analysis in Brummelman et al. (2015)

An investigation showing similar longitudinal patterns for boys and girls, helping rule out gender as a third-variable explanation.

Ethical Constraints on Praise Experiments

The ethical limitation preventing researchers from randomly assigning children to prolonged, potentially harmful overpraise for long durations.

Effort/Strategy Praise vs. Overvaluation Experiments

Short, ethically reviewed praise experiments demonstrating that praise directed at effort or strategy can produce favorable responses in children.

Statistical Significance Threshold

A statistical standard generally defined by $p < .05$ or a $95\%$ confidence interval (CI) that excludes zero.

n.s. (Not Statistically Significant)

A notation indicating that an association is not statistically significant, generally corresponding to $p > .05$ or a $95\%$ confidence interval including zero.

Multiple Regression

A statistical technique that examines whether a key relationship remains after accounting for other measured variables (controlling for them).

Criterion Variable (Dependent Variable)

The outcome variable being predicted in a multiple regression model, typically limited to one per regression analysis.

Predictor Variable (Independent Variable)

A variable used to predict the criterion outcome; multiple regression incorporates two or more predictor variables.

Controlling for a Third Variable

Asking whether a predictor is still linked to the criterion outcome when statistical variation associated with a third variable ($X$) is taken into account.

Statistical Control: Association Remains

A regression outcome indicating that the tested third variable does not fully explain the original association.

Statistical Control: Association Disappears

A regression outcome indicating that the tested third variable might explain the original association.

Chandra et al. (2008) Sample and Design

A study of $1{,}461$ teenagers who reported viewing of $23$ programs ($14$ episodes coded for sexual content per program), with pregnancy reported $3$ years later.

Chandra et al. (2008) Core Finding

The finding that higher exposure to sexual TV content was associated with higher pregnancy risk even after controlling for several third variables.

Measured Third Variables in Chandra et al. (2008)

Extraneous variables accounted for in the regression model, including age, parental monitoring, total TV time, grades, and family structure.

Age Control Logic in Chandra et al. (2008)

The principle that if the TV-pregnancy association remains among teens of the same age, age alone does not explain the relationship.

Beta ($\beta$)

A standardized regression coefficient indicating the direction and strength of a predictor's relationship with the criterion, controlling for all other predictors in the model.

Positive Beta ($\beta$)

A standardized regression coefficient indicating that higher levels of the predictor are associated with higher levels of the criterion outcome, controlling for other predictors.

Negative Beta ($\beta$)

A standardized regression coefficient indicating that higher levels of the predictor are associated with lower levels of the criterion outcome, controlling for other predictors.

Near-Zero Beta ($\beta$)

A standardized regression coefficient indicating little to no association with the criterion after controlling for other predictors.

Comparing Betas Within the Same Model

The practice of comparing standardized $\beta$ values within the same regression model or table to determine which predictor has a stronger relationship with the criterion.

Comparing Betas Across Different Regression Models

An inappropriate practice, because the specific set of controlled variables changes the meaning of each $\beta$.

Unstandardized Coefficient ($b$)

A regression coefficient that preserves predictors' original units; its sign shows direction, but raw sizes across different units cannot be compared to determine relative strength.

95% Confidence Interval in Multiple Regression

A range conveying the precision and size of an estimated regression effect; excluding zero indicates statistical significance, whereas including zero means not significant.

Chandra et al. (2008) Illustrative Two-Predictor Model

A textbook illustration where TV content ($\beta = .25$, $95\%$ CI $[.14, .36]$) and age ($\beta = .33$, $95\%$ CI $[.20, .46]$) were both positive and significant.

Chandra et al. (2008) Multivariable TV Content Coefficient

The reported multivariable regression coefficient of $0.44$ for sexual TV content, which remained statistically significant after several predictors were included.

Family Meals and Academic Achievement Study

An example where an initial bivariate correlation between family meals and school performance disappeared after controlling for parental involvement and other factors.

Family Meals Beta After Controlling for Parental Involvement

The illustrative regression coefficient for family meals, which dropped to $\beta \approx -.01$, $95\%$ CI $[-.06, .03]$ (not statistically significant).

Parental Involvement Beta in Academic Achievement Study

The illustrative regression coefficient for parental involvement predicting grades after controls, remaining statistically significant at $\beta \approx .09$, $95\%$ CI $[.06, .12]$.

Two Primary Functions of Multiple Regression

Testing third-variable explanations and identifying which additional variables predict the criterion outcome.

Critical Limit of Multiple Regression (Unmeasured Variables)

The inherent limitation that regression can only control for variables that researchers actually measured, leaving unmeasured factors potentially responsible.

Temporal Precedence Limit of Multiple Regression

The limitation that multiple regression alone may not establish the time order of variables.

Journalistic Indicators of Multiple Regression

Media phrasing such as 'controlled for,' 'adjusted for,' 'even after considering,' or 'accounting for' that indicates multiple regression was performed.

Randomized Experiments vs. Multiple Regression Control

The causal advantage of experiments because random assignment balances both measured and unmeasured third variables, whereas regression only controls measured variables.

Pattern and Parsimony

An approach using results from multiple studies and methods pointing in the same causal direction explained by the simplest theory with the fewest assumptions.

Pattern (in Pattern and Parsimony)

The consistency where findings from diverse studies, samples, and methodological designs all point toward the same causal direction.

Parsimony

The scientific preference for the simplest theory that explains the entire pattern of findings with the fewest exceptions or added assumptions.

Weight of Evidence (Convergence of Evidence)

The scientific confidence achieved when many independent lines of research support a claim, making it far more persuasive than an isolated correlation.

Smoking and Lung Cancer Initial Correlation

The initial bivariate relationship between smoking and lung cancer, reported with an estimated correlation of $r = .49$, $95\%$ CI $[.39, .58]$.

Biological Mechanism of Smoking and Lung Cancer

The explanatory principle that toxic chemicals in tobacco smoke directly damage human tissue, causing cancer risk to rise with exposure.

Five Converging Patterns in Smoking Research

The diverse empirical findings (duration, quitting, cancer localization, filter type, secondhand smoke) collectively explained by exposure to cigarette toxins.

Smoking Duration Finding

The empirical finding that a longer history of cigarette smoking predicts a higher risk of developing lung cancer.

Smoking Cessation Finding

The empirical finding that quitting smoking leads to a lower risk of lung cancer than continuing to smoke.

Localization of Lung Cancer Finding

The empirical finding that cancers among smokers concentrate specifically in the lungs and in particular tissue types exposed to smoke.

Filtered vs. Unfiltered Cigarettes Finding

The empirical finding that smokers of filtered versus unfiltered cigarettes exhibit different levels of cancer risk.

Secondhand Smoke Finding

The empirical finding that passive, secondhand exposure to tobacco smoke also increases cancer risk.

Failure of Third-Variable Accounts for Smoking Patterns

The inability of proposed third variables (such as coffee drinking) to explain all five diverse empirical patterns simultaneously.

Theory-Data Cycle in Pattern and Parsimony

The process where a theory leads to specific predictions, studies test those predictions, and consistent findings progressively strengthen the theory.

TV Violence and Aggression Research Evidence

A converging body of correlational, experimental, child, adult, and longitudinal research demonstrating that TV violence links to slightly more aggressive behavior.

Caution Regarding Single-Study Media Reports

The guideline that news reports highlighting a single new study can misrepresent scientific evidence, emphasizing the need to consult previous research and replications.

Mediator (Mediating Variable)

An intermediate variable or mechanism proposed to explain how or why one variable relates to an outcome ($A \rightarrow \text{mediator} \rightarrow B$).

Structural Sequence of Mediation

The causal pathway where variable $A$ leads to the mediator, which in turn leads to outcome $B$ ($A \rightarrow \text{mediator} \rightarrow B$).

Deep Conversations and Well-Being Mediation Model

A mediation model proposing that deep conversations lead to stronger social ties, which in turn lead to better subjective well-being.

Conscientiousness and Health Mediation Model

A mediation model proposing that conscientiousness leads to better compliance with medical advice, which subsequently leads to better physical health.

Statistical Methods for Testing Mediation

The statistical techniques, specifically multiple regression or structural equation modeling (SEM), used to estimate whether data fit a proposed mediation model.

Temporal Precedence Requirement in Mediation

The requirement that variable $A$ must be measured or manipulated first, followed by the mediator, and then outcome $B$, because simultaneous correlation cannot establish mediation.

Third-Variable Explanation (Confounder)

An external variable that accounts for both original variables, challenging internal validity by explaining away their relationship.

Moderator Variable

A variable that changes the strength or direction of an association between two variables, addressing questions of 'when' or 'for whom' a relationship exists.

Mediator vs. Third-Variable Explanation Distinction

A mediator is an internal causal step between cause and effect, whereas a third-variable explanation is an external factor causing both original variables.

Mediator vs. Moderator Distinction

A mediator explains 'why' or 'how' an association occurs, whereas a moderator explains 'when' or 'for whom' the association changes.

Sexual TV Comfort Mediational Hypothesis

A proposed mediational path where sexual TV content increases pregnancy risk by first increasing psychological comfort with sexual behavior.

Construct Validity in Multivariate Studies

An evaluation of how accurately, reliably, and appropriately variables are operationalized and measured, including questionnaires and coding protocols.

Interrater Reliability in Content Coding

The degree of agreement between independent coders who assess scenes or behaviors, such as coding sexual content in television episodes.

External Validity in Brummelman et al. (2015)

The question of generalizability in a sample of $565$ children from $17$ Dutch schools ($75\%$ participation rate), where voluntary participation limits certainty about representativeness.

External Validity in Chandra et al. (2008)

The question of generalizability across a geographically broad U.S. teen sample resembling national demographics, though without reported random selection.

Statistical Validity in Multivariate Studies

The assessment of effect sizes ($r$, $\beta$), precision ($95\%$ CIs), statistical significance, replication, outliers, and potential curvilinear relationships.

Smartphone-Checking Daily-Diary Study (Cognitive Failures)

A $7\text{-day}$ daily-diary study of $181$ iPhone users showing that more daily checks predicted more daily cognitive failures after controlling for demographics, stressors, and affect.