A. True Experiment Purpose??? Key to achieving???.

A. True Experiment

• Purpose???

• Key to achieving???

B. Quasi-experiment

Is a treatment but the researcher often does not introduce the intervention or treatment.

e.g. ---Hazards research

The assignment of subjects to treatment conditions is not random

e.g.---intact classrooms

X

Internal Validity Threats in Quasi-Experimental

Designs

• Group threat of Nonequivalence

• example of Head Start programs

-----Subtype-----Single Group Interrupted Time

Series design

X

•Is there an intervention effect???

•Problem---Time threats to internal validity

Time--------------------------->

Examples of single group interrupted time series designs

Single district studies of impacts of implementation of school desegregation plans on:

–student achievement

–white enrollment declines

Strategy for handling time threats in interrupted time series

Multiple group interrupted time series designs

X

control factor

C. Correlational or EX POST FACTO Designs

•Researcher has no control over either assignment or treatment

•Starts with effect and works backwards in time to identify causes.

Most common design type in quantitative social research

Simple example of a Correlational Design

Education Voted Did notVote

A Lot 70 30

A Little 30 70

Use of Correlational Design to Identify Unusual Cases

--More Complex Example--

Late 20th C Immigration and Internal Migration Patterns for

U.S. Metro. Areas

1985-1990 Migration Components

Metro area Immigration Internalmigration

Los Angeles 899,007 -174,673New York 756,034 -1,065,580San Francisco 293,306 -103,498Miami 210,609 45,287Chicago 179,524 -293,185

TRUE EXPER. ---- QUASI -----CORRELATIONAL

more <------------CONTROL----------------------less

result

more <--------INTERNAL VALIDITY-------- less

The EXTERNAL VALIDITY issue

(generalizability)

Threats to generalizability:

•nonrepresentative subjects

•put in artificial settings

•submitted to artificial trreatments

TRUE EXPER. ---- QUASI -----CORRELATIONAL

more <------------ARTIFICIALITY------------------less

result

less -----------EXTERNAL VALIDITY--------> more