Confounding Variables and Research Design Principles

Confounding Variable

A confounding variable is an outside factor that varies with the independent variable and could be causing the effect you see in the dependent variable.

Threats to Internal Validity

Factors that can affect the validity of a study's results, including history, maturation, regression to the mean, testing effects, selection threat, and attrition threat.

History

Events outside the study happen between pre- and post-test.

Maturation

Natural changes in participants over time.

Regression to the Mean

Extreme scores tend to move toward average on a retest.

Testing Effects

Taking a test influences performance on a later test.

Selection Threat

Groups are not equivalent at the start.

Attrition Threat

Participants drop out of the study, especially in one condition.

Experimenter Bias

The experimenter might unknowingly influence results (e.g., by giving cues or interpreting behavior differently).

Reasons for Null Results

Not enough between-group difference, too much within-group variability, measurement issues, ceiling/floor effects, weak manipulation or small sample.

Factorial Design

A study design that includes more than one independent variable.

Factor

Another name for an independent variable.

Level

The different values or groups within a factor.

2x2 Design

A design that means 2 factors, each with 2 levels.

Between-Subjects Design

Different people in each condition.

Within-Subjects Design

Same people do all conditions.

Mixed Design

One factor is between, one is within.

Main Effect

The overall effect of one independent variable, ignoring the other.

Interaction

The effect of one independent variable depends on the level of another.

Parallel Lines in Graphs

Indicate no interaction.

Crossing Lines in Graphs

Indicate interaction.

Pearson's r

A measure that tells you the strength (from -1 to +1) and direction (positive or negative) of a correlation.

Example of Strong Negative Correlation

r = -.80 means a strong negative correlation.

Scatterplots

Visual way to show the correlation.

Correlation Strength

Closer the dots to a straight line, the stronger the correlation.

Construct Validity

Are you measuring what you meant to?

Statistical Validity

Is the correlation statistically significant? How big is r?

Internal Validity

Are there confounds or third variables?

External Validity

Can it generalize to other people/situations?

Correlation ≠ Causation

You need covariance, temporal precedence, and no confounds.

Slope of Regression Line

Shows predicted change in Y for each unit increase in X.

Bivariate Correlation Limitation

It doesn't establish temporal precedence or rule out third variables.

Beta Coefficient

Tells you how strong and what direction a predictor variable has on the outcome (Y), controlling for others.

Reading a Beta Table

Look for significance (p-values) and size of beta.

Cross-Lagged Panel Design

Measures variables at multiple time points to see which came first.

Partial Correlation

Controls for third variables to get cleaner results.

Moderator

Changes the strength/direction of the effect.

Mediator

Explains how or why a relationship happens.

Third Variable

Another variable causing both variables.

Quasi-Experimental Design

Looks like an experiment, but no random assignment.

Non-equivalent Control Group

Groups that aren't randomly assigned.

Interrupted Time Series

Measuring outcome over time before/after event.

Field Experiments

Done in real-world settings with some control.

Small-N Designs

Focus on individual data over time.

Stable Baseline

Measure for a while, then introduce IV.

Multiple Baseline

Introduce treatment at different times across people or settings.

Reversal/Withdrawal (ABA)

Introduce, remove, and reintroduce IV to see effect.

Validities in Small-N vs. Large-N

Strong internal validity but limited external validity.

Good Theory Characteristics

Falsifiable, parsimonious, supported by data, generative.