final for research methods

lets get this effing bread

Between subjects (independent samples)

Different people experience different levels of the IV. Each person experiences one level

Establishing causation requires:

• Correlation / covariance

• Directionality / temporal precedence

• Internal validity

Covariance

If two variables arent correlated, they aren't causally related

Temporal precendence

Cause must come before the effect

How can a researcher enforce precedence?

By manipulating a variable

Internal validity

Eliminating other options by using experimental controls and equivalent groups

Independent variable

The cause under investigation; what the experimenter manipulates

Conditions

2 or more levels of the IV (control condition vs experimental/treatment condition)

Dependent variable

The effect: what the experimenter measures

Control variables

Variables that are held constant in the study

Extraneous variables

Variable that is not controlled or manipulated

Confounding variable (AKA confound)

A variable that varies systematically with the IV and can explain the results

What is experimental control?

Ways for an experimenter to eliminate confounding

We want everything between conditions to be the same EXCEPT

The level of our IV

Sampling

How we select people from the population

Assignment

How we place that sample into the conditions of the experiment (IV levels)

Sample random assignment

Any differences between individuals should be equally spread across conditions by chance

Sample random assignment experiment method

For each participant from the sample, randomly assign them to a condition

Matched random assignment

A tightly controlled way to cancel out differences in a potential confounding variable

Matched random assignment methods

If you have identified a possible confounding variable, match the participants on that confounding variable across the levels of the IV

Internal validity (longer definition)

The extent to which a research study produces a single explanation for the relationship between two variables

The goal of the experiment is to…

… say that it is manipulation of the IV that causes a change in the DV

External validity

When the results from one study can be replicated or generalized to other samples, research settings, and procedures

Population

Other participants, cultures, genders, ages, etc.

Ecological

From lab situation to real world

Temporal

To other periods of time (of the day, of the year) or generations

Higher internal validity leads to…

… lower external validity (opt for internal over external)

Within-subjects (AKA repeated measures design)

Each participant experiences ALL levels of the IV and are tested after each

Advantages of within-subjects design

• No worries about group equivalence

• Statistically more powerful (detect smaller differences with more confidence)

• Need fewer participants

Disadvantages of within-subjects design

• Practice

• Fatigue

• Sensitization

• Carryover effects

• Order effects

Practice effects

Getting better by repetition

Fatigue effects

Get tired or bored by repetition

Sensitization

By being repeatedly exposed, it makes the purpose of the experiment clear

Carryover effects

One level of the IV still effects you going into the second variable

Order effects

Participants behavior is affected by their order of the conditions

Quasi-experimental

Instead of manipulating IV, just study people who already express different levels of IV

Large individual differences could be confounds so…

… use within-subjects

Long carryover effects or order effects?

Use between subjects

Hard to recruit participants? Use…

… within-subjects

Counterbalancing design

Presenting levels of the IV in different orders to different participants

Confounding variables overview

If you DO find a significant effect of your IV on your DV, consider these to rule out alternative explanations

Obscuring variables

If you DO NOT find a significant effect of your IV on your DV, consider whether these prevented you from seeing an effect that truly exists

One group pretest-posttest design

A single group is measured before and after an intervention to look for change

Maturation threats

Group behavior may change due to the passage of time

How to rule out maturation threats?

Control group with no manipulation of IV critical

History threat

Observed changes are due to some other uncontrolled event of change in the enviornment that has happened

How to rule out history threats?

No-treatment IV group

Regression to the mean

Group with extreme score is more likely to score closer to the mean (less extreme) at another time point.

Attrition threat

When participants with a particular characteristic drop out systematically from your study

How can attrition threats be ruled out/prevented?

• Look at pre-test scores and see if those who dropped out were different in any consistent way that those who did not

• Can remove their pre-test scores, but then there's concern about external validity

Observer bias

Observer seeing what they want to see

Demand characteristics

When participants think they know what experiment is about and change their behavior

Placebo effects

A participant's belief that they are receiving an effect treatment that leads to actual improvement

How to prevent threats to internal validity?

• Single- or double-blind studies

• Clear operational definitions

Possible explanations for null effects

• Not enough between-groups difference / systematic variance

• Too much within-groups variability / error variance

• No actual difference / no effect of your IV on DV

Not enough between-groups difference / systematic variance causes

• Weak manipulations

• Insensitive measures

• Ceiling and floor effects

• Use manipulations checks

Too much within-groups variability / error variance causes

• Measurement errors

• Individual differences

• Situation noise

Power

The likelihood that a study will yield a statistically significant result when the IV really has an effect

Studies with a lot of power are

More likley ro detect true differences