Statisitcs: Final Exam Study Guide

Convenience Sample

The data is collected from an easily accessible and available group of people

Key Features of an Experiment

• Manipulation of the dependent variable

• Measurement of the independent variable

• Random assignment

• Control

Manipulation of the Dependent Variable

Needs to be sufficient to have a measurable impact on the dependent variable

Measurement of the Independent Variable

Manipulation checks measure the impact of the independent variable

Random Assignment

It is truly random which condition participants are assigned to

Experimental Control

No confounds

Why an Experiment Enables us to Establish Cause and Effect Relationships

Manipulation

Between Subjects Design

Different group of subjects for each level of the independent variable

Within Subjects Design

Each subject experiences every level of the independent variable, and is remeasured after each level

Equality: Within Subjects

• The same person in both groups

• Counterbalancing

Counterbalancing

• Order of manipulations is varied over subjects

• Any differences over time should average out, re equalizing groups

Equality: Between Subjects

• Random assignment

Random Assignment

• Participants are assigned to groups randomly

• This should balance out any differences and equalize the groups

Matching Assignment

Participants are pretest, matched, and then randomly assigned

Carryover

Manipulation has long lasting effects

Carryover and Counterbalancing

Has different participants exposed to each condition in a different order so that the order of conditions does not systematically favor any particular treatment

Internal Validity

Quality of the internal logic of the experiment

How to Achieve Internal Validity

• Achieved experimental control, no confounds

• Achieved strong and effective manipulation of the independent variable

• Achieved reliable and valid measurement of the dependent variable

• Dependent variable was sensitive to possible changes in the independent varible

Threats to Internal Validity

• Confounds

• Failure to conduct true random assignment

• Differential dropout

• Experimenter behavior

• Participant expectations

• Placebo effect

• Failure to manipulate the independent variable

Threats to Internal Validity: Confound

A variable that varies systematically with the independent variable, providing an alternative interpretation

Threats to Internal Validity: Experimenter Behavior

Treating groups differently

Threats to Internal Validity: Participant Expectations

Different expectations can lead to different behaviors, regardless of treatment

External Validitiy

How well the experiment generalizes to other contexts

How to Achieve External Validity

• Different ways of operationalizing the independent variable

• Different ways of operationalizing the dependent variable

• Different participants

• Different contexts

• In the real world

Threats to External Validity

• Unable to replicate the study

• Replication does not match original study

• Selection bias

Ecological Validity

In the real world

Features of a Factorial Design

Having more than one independent variable, with levels

Cohen’s D: No Effects

0 to 0.2

Cohen’s D: Small Effects

0.2 to 0.5

Cohen’s D: Moderate Effects

0.5 to 0.8

Cohen’s D: Large Effects

0.8 and more

The Mean Difference

The effect size, usually Posttest - Pretest

Cohen’s D

The difference between two means divided by a standard deviation for the data, and it measures the differences between the two means

Null Hypothesis

A statement of no effect or no relationship

Research Hypothesis

A statement that introduces a research question and proposes an expected result

Type I Error

Assuming there’s an effect, but there might not really be one

Type II Error

Assuming there’s not an effect, but there might actually be one

If P is Less Than 0.05

• The finding is statistically significant

• Reject the null hypothesis

• The confidence interval will not include zero

• T is likely at least 2 or -2

• At risk of Type I Error

If P is Greater Than or Equal to 0.05

• Not statistically significant

• Fail to reject the null

• The confidence interval will include zero

• T will be something less than 2

• At risk of Type II error