Textbook Notes (280,000)
CA (170,000)
U of G (10,000)
MCS (700)
Chapter 8

# MCS 3030 Chapter Notes - Chapter 8: Internal Validity

Department
Marketing and Consumer Studies
Course Code
MCS 3030
Professor
Tanya Mark
Chapter
8

This preview shows half of the first page. to view the full 3 pages of the document.
Chapter 8 Research Methods
Experimental Design
8-1 INTRODUCTION TO EXPERIMENTAL DESIGN
8-1A Experimental Designs & Internal Validity
- Experimental designs are usually considered the strongest of all designs in internal validity b/c internal
validity is at the center of all causal or cause-effect inferences.
- If you are able to provide evidence for both propositions, (If X, then Y & If not X, then not Y), than you’ve in
effect isolated the program from all the other potential causes of the outcome
- In the simplest type of experiment, you create 2 groups that are equal to eachother. One group gets the
program, and the other doesn’t. In all other respects the groups are treated the same. Now if you see differences
in outcomes between the 2 groups, they must be due to the only thing that differs between them… the program.
- Key to success of the experiment is random assignment of people into groups.
8-1B Two-Group Experimental Design
- The simplest form of experimental design
- In design notation, it has 2 lines one for each group with an R at the beginning of each line to indicate that
the groups were randomly assigned
1 group gets treatments/program (the X) and the other doesn’t
- A pretest isn’t required for this design b/c this uses random assignment and you can assume that the 2 groups
are probabilistically equivalent to begin with
- Most interested in this design in determining whether the 2 groups are different after the program.
- Typically you measure the groups on one or more measures (the Os in notation), and you compare them by
testing for the differences between the means using a t-test or one-way analysis of variance (ANOVA)
ANOVA: An analysis that estimates the difference between groups on a posttest.. The ANOVA could estimate
the difference between a treatment and control group (thus being equivalent to a t-test) or can examine both
main and interaction effects in a factorial design.
Selection-Mortality Threat: A threat to internal validity that arises when there is differential nonrandom
dropout between groups during the study.
- B/c the design requires random assignment, in some settings such as schools, it is more likely to utilize
persons who would be aware of eachother and of all the conditions to which you have assigned them.
8-1C Probabilistic Equivalence
- The notion that 2 groups, if measured infinitely, would on average perform identically. These groups would
seldom obtain the exact same average score in a real setting.
- You know perfectly the odds of finding a difference between the 2 groups
- Can achieve probabilistic Equivalence through random assignment to groups b/c you can calculate the chance
that the 2 groups will differ
8-1D Random Selection & Assignment
- Random selection is how you draw the sample of people for your study from a population
Related to sampling therefore it is most closely related to external validity of your results
Random Assignment: How you assign the sample that you draw to different groups/treatments in your study
Most closely related to design and therefore most closely related to internal validity
8-2 CLASSIFYING EXPERIMENTAL DESIGNS
- Assume that what you observe in a research study can be divided into 2 components: the signal & the noise
Variability (Noise): The extent to which the values measured or observed for a variable differ
- Inmost research, the signal is related to the key variable of interest the construct you’re trying to measure –
or the program/treatment being implemented
- The noise consists of all the random factors in the situation that make it harder to see the signal: lighting in the
room, how people feel that day, distractions etc.
- Can construct a ratio of these 2 by dividing the signal by the noise
In research you want the signal to be high relative to the noise
Can divide experimental designs into 2 categories: signal enhancers & noise reducers
1. Factorial Designs (Signal Enhancing): Designs that focus on the program/treatment, its components, and