SingleFactor Experiments What is a true experiment Betweensubjects

Single-Factor Experiments • • What is a true experiment? Between-subjects designs Within-subjects designs Designs to avoid (not true experiments)

What is a true experiment? Experiment = study in which researcher has complete control over all aspects of the study 2 Essential features of experiments: • Control Group or Control Condition (actually, or 2 or more levels of an IV) • Random allocation of subjects to groups (if it has a between-subjects factor)

Some Terminology • IV = what the experimenter manipulates (varies) in an experiment; the hypothesized cause • DV = what the experimenter measures to test the hypothesis in an experiment; the hypothesized effect • Factor = IV • Level = condition = treatment: – One value of an IV – Example: for the factor “gender, ” the levels are “male” and “female” (in a quasi-experiment) – Example: for the factor “number of witnesses” in a bystander intervention experiment, the levels could be “one” “two” and “four”

Between-subjects designs • At least 2 conditions (groups) – Control and Experimental Condition, or – 2 or more levels of IV without “control” condition • Each subject is assigned to only one condition • Random assignment of subjects to conditions • Design a between-subjects experiment testing this hypothesis: Vocalization impairs short-term memory.

Between-subjects Designs with Multiple Conditions • More than 2 levels of the IV • Randomly assign subjects to conditions Design a between-subjects experiment testing to answer the following questions: – Does vocalization impair STM? – Does it matter whether the vocalization is aloud or just “in your head”?

Within-Subjects Designs • At least 2 conditions – Control and Experimental Condition, or – 2 or more levels of IV without “control” condition • Each subject is assigned to all conditions • Counterbalance to control for order and sequence effects Design a within-subjects experiment testing this hypothesis: Sub-vocalized speech impairs short-term memory.

Counterbalancing in Within-Subjects Designs • Counterbalancing within subjects – Useful when each level of the IV occurs multiple times for each subject – Each subject gets both sequences of the conditions • Counterbalancing within groups (not within each subject) – Useful when there are many (more than 2) conditions – Necessary if each condition occurs only once per subject – An equal number of subjects get each sequence of the conditions

Order and Sequence Effects • Order effects – Result from the position in which a condition occurs (first, second, third, etc) – If the order of conditions is “A-B”, order effects on B would be the effects that result from B being in the second position in the list – Example: practice effects • Sequence effects – Result from which conditions precede or follow a condition – If the order of conditions is “A-B”, sequence effects on B would be the effects that result from B following A

Controlling Order and Sequence Effects • Controlling Order Effects – Counterbalance the position in which each condition appears – ABC, CBA, ACB: C occurs first, second, and third an equal number of times • Controlling Sequence Effects – Counterbalance what each condition follows – ABC, BAC : C follows A half the time, and follows B half the time.

Controlling Order and Sequence Effects Within Subjects • Only possible when each condition occurs at least twice for each subject • Method depends on how many times each condition occurs per subject: – Many times: randomize order of conditions – A few times: use block randomization of conditions – Twice: use reverse counterbalancing

Controlling Order and Sequence Effects Within Groups • When it is not possible or practical to control within subjects • Necessary if each condition occurs only once for each subject • Sequences and orders controlled within a group of subjects, but not within each subject • Latin Square – a technique for partial counterbalancing (when full counterbalancing is impractical)

Latin Square • A square matrix with length = number of conditions • Each condition occurs only once on each row and only once in each column: abcd bcda cdab dabc

Balanced Latin Square • A Latin Square in which each condition is preceded by every other condition exactly once: abcd bdac cadb dcba

Designs to Avoid • One-group post-test only • Post-test only with non-equivalent control group • One-group pretest-posttest design