APEX AP Stats 4.1.4 Vocabulary of Experiments and Surveys Study Guide

Experiment

Method of collecting data in which the researcher exerts control over the subjects or units in the study.

What are experiments done to establish?

cause and effect

What do we want in an experiment?

a sufficient number of subjects/units in the experiment so we can be sure the results aren't due to chance (Chance variation)

Treatments (Factors)

Applied interventions in an experiment to measure responses.

Chance Variation

Variation in results that may occur due to random chance.

Replication

Repeating the experiment to reduce the probability of chance variation.

What can replication do?

reduces error due to chance and can reduce bias

Experimental Units

Things that are being experimented on.

Subjects

Human experimental units.

Explanatory Variable (Factor)

Independent variable that tries to explain the response variable; the treatments for which we measure a response from.

Two Factor Experiment

An experiment where multiple things are being tested.

Levels of the Factor

The amount of the factor that is being used for the treatments.

Response Variable

Dependent variable that happens because of the explanatory variable.

Bias

Error that encourages one outcome over another.

Placebo

Fake treatments.

Placebo Effect

An effect that happens when subjects believe they're receiving the treatment but aren't and they experience an improvement/effect.

Blind

Subjects don't know if they're receiving the treatment or placebo.

Double-Blind

Both subjects and administrators of the treatments are unaware of who is receiving the treatment or placebo.

Three Primary Principles of Experimental Design

Replication, Control, Randomization.

Generalizability

The goal of an experiment/sample survey; applying findings to the larger population our treatment groups represent.

Random Sampling

Necessary to ensure treatment group is free of bias.

Lurking Variable

Variables that you aren't aware of but may be causing the response.

Confounding Variables

Another term for lurking variables.

Control Group

A group that is kept identical to the treatment group to prevent confounding variables.

What does control ensure?

differences are caused by the independent variable

Treatment Group

The group that receives the treatment in an experiment.

Confounded

We can't tell which variable might be causing an effect.

Hawthorne Effect

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

Comparative Study

A form of an experiment that compares two different variables to see which one has the most effect.

What should be present in the results of a comparative study?

Any differences will be meaningful if the groups are similar in the beginning and If the placebo effect and Hawthorne effect are present, they should be the same for both groups

Control Idea

An experiment's confounding variables should operate equally on both groups.

Matched Pairs

Subjects in the treatment and control groups are matched on some quality or qualities that make them all alike as possible.

Are matched pairs a special case of blocking?

yes

What are the common ways to create matched pairs?

1. Match groups on a list of factors and use one group as the treatment group and another group as the control group 2. Before and after measurements on the same individuals

Response Variables

Variables that might contribute to the response and should be matched.

Gain Score

The difference between the before and after measurements for each individual.

Response Bias

Bias caused by variables that may affect the control group and treatment group differently.

What do matching pairs prevent?

response bias

Randomization

Researchers randomly choose subjects from the population and randomly place them in the treatment and control group.

What does randomization ensure?

1. the effects of working variables are canceled out by the two groups 2. allows us to ensure that the difference between the groups are caused by the independent variable

When is blocking done in an experiment?

before randomizing when there are blocks within the subjects and if we expect some variables to have an important effect of their own

What are the two ways to randomize a group?

1. Generate random numbers 2. Use a table of random numbers

Blocking

Separating experimental units because of potential variations that may affect the results of the experiment.

What is needed to generalize the results of an experiment to a larger population?

if we use experimental control and replicate the experiment and get the same thing

Control

Account for lurking variables to ensure that differences are caused by the independent variable.

Chance Variable

Variability due to chance that should be minimized.

Experimental Control

Using control groups and treatment groups to generalize results to a larger population.

Lurking Variables

Variables that may affect the outcome but are not accounted for in the study.

Effects of Working Variables

Ensured to be canceled out by randomization.

Observational Studies

Studies that can use matched pairs to link treatment and control groups.

Experimental Studies

Studies that can use matched pairs to link treatment and control groups.

Before and After Measurements

Measurements taken on the same individuals to assess effects.

Randomly Assigned

Experimental units are randomly assigned to treatment and control groups.