Research Methods in Psychology Chapter 9

Regression in Popular Press Articles

"controlled for," "taking into account," and "correcting/adjusting for"

Cross-Lag Correlations

(researchers are most interested in these) they show whether the earlier measure of one variable is associated with the later measure of the other variable. EXAMPLES: TV violence in the 3rd grade --> Aggression in the 13th grade Aggression in the 3rd grade --> TV violence in the 13th grade

Differences between Mediators and 3rd Variables

3rd variables are external to the bivariate correlation (problematic) Mediators are internal to the causal variable (not problematic)

The Power of Pattern and Parsimony

EXAMPLE: does smoking cause lung cancer? problems: endless amount of 3rd variables and smoking experiment is not ethical or practical BUT: simple theory that "the more contact human tissue has with these chemicals, the more toxicity they are exposed to" leads to the parsimonious theory that chemicals in cigarettes cause cancer

Controlling for 3rd Variables

EXAMPLE: minutes for recess may affect behavior problems, but income level may affect both behavior problems and minutes for recess

3rd Variable Problem

Two variables are correlated, but only because they both linked to a third variable EXAMPLE: "The relationship between viewing violent TV and aggressive behavior may be attributable to the third variable of parental leniency."

Moderator

WHEN, for WHOM, under WHAT conditions are two variables related? EXAMPLE: "Gender moderates the relationship between TV violence and aggressive behavior."

Mediator

WHY are two variables related? EXAMPLE: "level of desensitization mediates the relationship between TV violence and aggressive behavior."

Similarities between Mediators and 3rd Variables

both involve multivariate research designs both can be detected using multiple regression

When PREDICTION is the goal you:

can throw in as many predictors as possible, as long as they contribute a positive predictive value to the outcome variable.

Applications of Regression Techniques

controlling for third variables (and so on), mediational analyses, path models and structural equation models

Three Criteria for Establishing Causation

covariance, temporal precedence, and internal validity

Criterion Variable

dependent variable; the variable they are most interested in understanding or predicting

Autocorrelations

determine the correlation of one variable with itself, measured on two different occasions EXAMPLE: TV violence in the 3rd grade --> TV violence in the 13th grade

Linear Regression

fits the "best" line to a scatter of points in two dimensions: Y= aX + b minimizes the sum of the squared vertical distance between all points and the line (sum of squared errors; least squares) predictive tool and most commonly used with TWO measured variables

Multiple Regression (Multivariate Regression)

helps rule out some 3rd variables and address questions of internal validity

Predictor Variable

independent variable; variables measured in a regression analysis

Multivariate Designs

involve more than 2 measured variables

Examples of Bivariate Designs

longitudinal, multiple-regression, and the pattern and parsimony approach

Mediation Analyses

looks at process oriented variables requires 3 steps: test that IV alone significantly predicts DV, tests that IV alone significantly predicts mediator, test that when your 2 predictors are the mediator and the IV, the mediator is a better predictor of the DV than the IV. EXAMPLE: recess availability (IV) leads to physical activity (mediator) which leads to behavior problems (DV)

Cross-Sectional Correlations

test to see whether two variables, measured at the same point in time, are correlated EXAMPLE: TV violence in the 3rd grade --> Aggression in the 3rd grade

Path Models and Structural Equation Models (SEM)

testing alternative models against each other and see which model fits the data the best; can also be used to confirm or reject a single theoretical model can get complicated can often see demographic variables and socioeconomic factors/outcomes EXAMPLE: models of the future purchase in intentions of customers (model of trust and commitment as mediators VS model of overall satisfaction as mediator)

Parsimony

the degree to which a good scientific theory provides the simplest explanation of some phenonmenon

When EXPLANATION is the goal you:

want to be very careful about adding more variable, even if they marginally increase variance accounted for in your data

Standardized Coefficients

what you need if your question is which variable contributes most of the predictive value (i.e. which predictor is most "important"

Unstandardized Coefficients

what you ultimately need to predict the outcome variable, if you interested in the actual value (like annual income, for example)