Research Methods and Statistics

Pretest

A measurement taken before the treatment to see where participants start.

Manipulated

Deliberately changed or controlled by the researcher.

Posttest

A measurement taken after the treatment to see if anything changed.

Independent variable

What the researcher manipulates or studies as the possible cause.

Dependent variable

What is measured; its value depends on the independent variable.

Depended

Varied in response to something else.

Free-recall task

Writing down everything you remember with no prompts or choices.

Credible

Believable and trustworthy.

Internal validity

How well a study's design and data support firm conclusions about cause and effect.

Warranted

Justified or supported by the evidence.

Hawthorne effect

People change their behavior because they know they're being studied.

Reactivity

A change in behavior simply because people know they're being observed.

Novelty effect

A change caused by something being new, which can fade over time.

Confounding variables

Outside factors that vary along with the treatment and muddy cause and effect.

Keep things constant

Making a factor the same for everyone so it can't explain differences.

Control group

A group that doesn't get the treatment, used for comparison.

External validity

How well findings generalize to other people, settings, and situations.

History

An outside event between pretest and posttest that changes the results.

Maturation

Change that happens just from time passing.

Testing

Better performance caused by practice from taking the test before.

Instrumentation

A change in how the measuring tool is used or scored.

Statistical regression

Extreme scores tend to move toward the average when measured again.

Selection

Groups differ from the start because of how people were chosen.

Attrition

Participants dropping out at different rates across groups.

Interrater reliability

How consistently different raters score the same thing.

Placebo

A fake treatment that looks real but shouldn't have an effect.

Double-blind experiment

Neither participants nor researchers know who is in which group.

Randomly assign people to groups

Using chance to place participants in groups so they are similar on average.

Informed consent

Participants agree to take part after being told what the study involves.

Pretests to assess equivalence

Testing groups beforehand to check they're similar on key variables.

All experimental treatments

Every participant gets every condition, acting as their own control.

Matched pairs

Pairing people in different groups who are alike on key traits.

Statistically control

Using statistics to account for outside factors you couldn't remove.

Partial correlation

A statistic that measures a relationship while removing a third variable's effect.

ANCOVA (analysis of covariance)

A test that compares groups while adjusting for a covariate.

Structural equation modeling

A statistical method testing how multiple variables relate in a model.

Pre-experimental designs

Simple designs with little control, useful for tentative hypotheses.

True experimental designs

Designs with random assignment and strong control.

Quasi-experimental designs

Designs used when randomization isn't possible, leaving some confounds.

Ex post facto designs

Studying a past 'cause' with no manipulation by the researcher.

Factorial designs

Designs that study two or more independent variables at once.

Reversal time-series design

Treatment is introduced, withdrawn, and reintroduced to check its effect.

Design 1: One-shot experimental case study

One group gets a treatment and is observed once.

Exp

A prior experience some participants had that the researcher didn't control.

Design 2: One-group pretest-posttest

One group is measured, gets a treatment, then is measured again.

Design 3: Static group comparison

A treated and untreated group are compared afterward, with no pretest.

Agronomist

A scientist who studies crops and soil.

Design 4: Control-group pretest-posttest

Randomized groups are measured before and after; only one gets treatment.

Design 5: Solomon four-group design

A four-group design checking whether pretesting itself affects results.

Experimental group

The group that receives the treatment.

Design 8: Nonrandomized control-group pretest-posttest

Like Design 4 but without random assignment.

Design 9: Simple time-series design

Many observations before and after an intervention to see change.

Randomness

Using chance to select or assign.

Multiple-case intervention research

Studying how an intervention affects both a group and individuals within it.

Multiple-baseline design

Treatment starts at different times for different participants to show causation.

Design 14: Simple ex post facto design

Comparing groups where one had an uncontrolled past experience.

Congenital

Present from birth.

Design 15: Two-factor experimental design

A design with two independent variables.

Moderating variable

A pre-existing characteristic changing how strongly a treatment works.

Mixed-methods design

Combining quantitative and qualitative approaches in one study.

Exhaustive

Covering every possibility.

Meta-analysis

A statistical combining of results from many studies.

Replication

Repeating a study to check that results hold up.

Conducting experiments on the Internet

Running experimental studies online.

Fluke

A lucky or unusual accident.

A priori hypotheses

Predictions made before collecting data.

Bias

A systematic influence pushing results away from the truth.

Type I error

Concluding there's an effect when there really isn't one.

Atypical

Not typical or normal.

Fetal monitor

A device that tracks an unborn baby's heart rate during labor.