Experimental Research

What is an Experiment?
Research method in which
conditions are controlled so that 1 or more independent variables can be manipulated to test a hypothesis about a dependent variable.

Allows
evaluation of causal relationships among variables while all other variables are eliminated or controlled.

Some Definitions
Dependent Variable
Criterion by which the results of the experiment are judged. Variable that is expected to be dependent on the manipulation of the independent variable

Independent Variable
Any variable that can be manipulated, or altered, independently of any other variable Hypothesized to be the causal influence

More Definitions
Experimental Treatments
Alternative manipulations of the independent variable being investigated

Experimental Group
Group of subjects exposed to the experimental treatment

Control Group
Group of subjects exposed to the control condition Not exposed to the experimental treatment

More Definitions
Test Unit
Entity whose responses to experimental treatments are being observed or measured

Randomization
Assignment of subjects and treatments to groups is based on chance Provides control by chance Random assignment allows the assumption that the groups are identical with respect to all variables except the experimental treatment

Constant Error (bias)

Constant error is error that occurs in the same experimental condition every time the basic experiment is repeated a systematic bias Example:
Experimental groups always administered the treatment in the morning Control groups always in the afternoon Introduces an uncontrolled extraneous variable time of day Hence, systematic or constant error

Extraneous Variables
Variables other than the manipulated variables that affect the results of the experiment Can potentially invalidate the results

Sources of Constant Error

Demand Characteristics
Experimental design procedures or situational aspects of the experiment that provide unintentional hints to subjects about the experimenters hypothesis If occurs, participants likely to act in a manner consistent with the experimental treatment. Most prominent demand characteristic is the person actually administering the experimental treatments.

Experimenter Bias
Effect on the subjects behavior caused by an experimenters presence, actions, or comments.

Guinea Pig Effect

Effect on experimental results caused by subjects changing normal behavior or attitudes to cooperate with experimenter.

Controlling Extraneous Variables

Blinding
Technique used to control subjects knowledge of whether or not they have been given the experimental treatment. Taste tests, placebos (chemically inert pills), etc.

Constancy of Conditions
Subjects in experimental & control groups are exposed to identical situations except for differing conditions of the independent variable.

Controlling Extraneous Variables

Order of Presentation
If experimental method requires that the same subjects be exposed to 2 or more experimental treatments, error may occur due to order in which the treatments are presented Counterbalancing
the subjects exposed to Treatment A first, then to Treatment B. Other exposed to Treatment B first, then to Treatment A. Eliminates the effects of order of presentation

Experimental Validity
Internal Validity
Indicates whether the independent variable was the sole cause of the change in the dependent variable

External Validity
Indicates the extent to which the results of the experiment are applicable to the real world

Extraneous Variables that Jeopardize Internal Validity

History Effect
Specific events in the external environment between the 1st & 2nd measurements that are beyond the experimenters control Common history effect occurs when competitors change their marketing strategies during a test marketing experiment

Cohort Effect
Change in the dependent variable that occurs because members of one experimental group experienced different historical situations than members of other experimental groups

Extraneous Variables that Jeopardize Internal Validity

Maturation Effect
Effect on experimental results caused by experimental subjects maturing or changing over time During a daylong experiment, subjects may grow hungry, tired, or bored

Testing Effect
In before-and-after studies, pretesting may sensitize subjects when taking a test for the 2nd time. May cause subjects to act differently than they would have if no pretest measures were taken

Extraneous Variables that Jeopardize Internal Validity

Instrumentation Effect
Caused by a change in the wording of questions, in interviewers, or in other procedures used to measure the dependent variable.

Selection Effect
Sampling bias that results from differential selection of respondents for the comparison groups.

Mortality or Sample Attrition

Results from the withdrawal of some subjects from the experiment before it is completed Effects randomization Especially troublesome if some withdraw from one treatment group and not from the others (or at least at different rates)

Figure 8.3 Experimentation as Conclusive Research Conclusive Research

Descriptive

Causal

Experimentation

Field Experiments

Laboratory Experiments

Laboratory Versus Field Experiments FACTOR Environment Control Reactive error Demand artifacts Internal validity External validity Time Number of units Ease of implementation Cost LABORATORY Artificial High High High High Low Short Small High Low FIELD Realistic Low Low Low Low High Long Large Low High

Symbolism for Diagramming Experimental Designs

X = exposure of a group to an experimental treatment O = observation or measurement of the dependent variable If multiple observations or measurements are taken, subscripts indicate temporal order I.e., O1, O2, etc. R = random assignment of test units; individuals selected as subjects for the experiment are randomly assigned to the experimental groups

Pre-Experimental Designs
Do not adequately control for the problems associated with loss of external or internal validity Cannot be classified as true experiments Often used in exploratory research Three Examples of Pre-Experimental Designs
One-Shot Design One-Group Pretest-Posttest Design Static Group Design

One-Shot Design
A.K.A. after-only design A single measure is recorded after the treatment is administered Study lacks any comparison or control of extraneous influences No measure of test units not exposed to the experimental treatment May be the only viable choice in taste tests Diagrammed as: X O1

One-Group Pretest-Posttest Design

Subjects in the experimental group are measured before and after the treatment is administered. No control group Offers comparison of the same individuals before and after the treatment (e.g., training) If time between 1st & 2nd measurements is extended, may suffer maturation Can also suffer from history, mortality, and testing effects Diagrammed as O1 X O2

Static Group Design

A.K.A., after-only design with control group Experimental group is measured after being exposed to the experimental treatment Control group is measured without having been exposed to the experimental treatment No pre-measure is taken Major weakness is lack of assurance that the groups were equal on variables of interest prior to the treatment Diagrammed as: Experimental Group X O1 Control Group O2

Pretest-Posttest Control Group Design

A.K.A., Before-After with Control True experimental design Experimental group tested before and after treatment exposure Control group tested at same two times without exposure to experimental treatment Includes random assignment to groups Effect of all extraneous variables assumed to be the same on both groups Do run the risk of a testing effect

Pretest-Posttest Control Group Design

Diagrammed as Experimental Group: R O1 X O2 Control Group: O3 O4 R Effect of the experimental treatment equals (O2 O1) -- (O4 O3) Example 20% brand awareness among subjects before an advertising treatment 35% in experimental group & 22% in control group after the treatment Treatment effect equals (0.35 0.20) (0.22 0.20) = 13%

Posttest-Only Control Group Design

A.K.A., After-Only with Control True experimental design Experimental group tested after treatment exposure Control group tested at same time without exposure to experimental treatment Includes random assignment to groups Effect of all extraneous variables assumed to be the same on both groups Do not run the risk of a testing effect Use in situations when cannot pretest

Posttest-Only Control Group Design

Diagrammed as Experimental Group: R X O1 R Control Group: O2 Effect of the experimental treatment equals (O2 O1) Example Assume you manufacture an athletes foot remedy Want to demonstrate your product is better than the competition Cant really pretest the effectiveness of the remedy

Solomon Four-Group Design

True experimental design Combines pretest-posttest with control group design and the posttest-only with control group design Provides means for controlling the interactive testing effect and other sources of extraneous variation Does include random assignment

Solomon Four-Group Design

Diagrammed as Experimental Group 1: Control Group 1: Experimental Group 2: Control Group 2: R R R R O1 O3 X X O2 O4 O5 O6

Effect of independent variable (O2 O4) & (O5 O6)

Effect of pretesting (O4 O6) Effect of pretesting & measuring (O2 O5) Effect of random assignment (O1 O3)

Quasi-Experimental Designs
More realistic than true experiments Researchers lacks full control over the scheduling of experimental treatments or They are unable to randomize Includes
Time Series Design Multiple Time Series Design
Same as Time Series Design except that a control group is added

Time Series Design

Involves periodic measurements on the dependent variable for a group of test units After multiple measurements, experimental treatment is administered (or occurs naturally) After the treatment, periodic measurements are continued in order to determine the treatment effect Diagrammed as: O 1 O2 O3 O4 X O5 O6 O7 O8

Statistical Designs
Multiple experiments are conducted simultaneously to permit extraneous variables to be statistically controlled and Effects of multiple independent variables to be measured Advantages
Can measure the effects of more than one independent variable Can statistically control specific extraneous variables Economical designs can be formulated when each subject is measured more than once.

Completely Randomized Design

Involves randomly assigning treatments to group members
Allows control over all extraneous treatments while manipulating the treatment variable Simple to administer, but should NOT be used unless test members are similar, and they are also alike regarding a particular extraneous variable Different forms of the independent variable are called levels.

Completely Randomized Design Example

Grocery store chain trying to motivate consumers to shop in their stores 3 possible sales promotional efforts X1 = offer discount of 5% off total shopping bill X2 = offer taste sample of selected foods X3 = control group, no sales promotional effort applied

Completely Randomized Design Example

SALES PROMOTION TECHNIQUE LEVELS 5% discount Sales, store 3 STORES Sales, store 1 Taste samples Sales, store 5 Sales, store 8 No sales promotion Sales, store 9 Sales, store 7

Sales, store 6
Average sales

Sales, store 4
Average sales

Sales, store 2
Average sales

Randomized Block Design

Randomly assigns treatments to experimental & control groups Test units broken into similar blocks (or groups) according to an extraneous variable
I.e., location, age, gender, income, education, etc.

Particularly useful when small sample sizes are necessary

Randomized Design Example

Grocery store chain trying to motivate consumers to shop in their stores 3 possible sales promotional efforts X1 = offer discount of 5% off total shopping bill X2 = offer taste sample of selected foods X3 = control group, no sales promotional effort applied Blocks = time stores have been in operation

Latin Square Design

Allows control or elimination of the effect of two extraneous variables Systematically blocks in 2 directions by grouping test units according to 2 extraneous variables Includes random assignment of treatments to each cell in the design Used for comparing t treatment levels in t rows and t columns
I.e., if we have 3 treatment levels, we must have 3 rows and 3 columns

Latin Square Design

Extraneous Variable 2 A B C

Extraneous Variable 1

B
C

C
A

A
B

where A, B, & C are all treatments

Latin Square Design Example

PER CAPITA INCOME TIME IN OPERATION High Medium Low

< 5 years

X1

X2

X3

5 10 years

X2

X3

X1

> 10 years

X3

X1

X2

Factorial Design
Used to examine the effects that the manipulation of at least 2 independent variables (simultaneously at different levels) has upon the dependent variable The impact that each independent variable has on the dependent variable is referred to as the main effect Dependent variable may also be impacted by the interaction of the independent variables. This is called the interaction effect

Factorial Design Example

Grocery store chain wants to use 12 of its stores to examine whether sales would change at 3 different hours of operation and 2 different types of sales promotions Dependent variable is change in sales Independent variables
Store open 6 am to 6 pm Store open 6 am to midnight Store open 24 hours/day Sales promotion: samples for a free gift Sales promotion: food samples

Called a 3 x 2 factorial design Need 6 experimental groups (3 x 2 = 6)

Factorial Design Example

HOURS OF OPERATION

SALES PROMOTION
Gift stamps

6 am 6 pm

5 am midnight

24 hours

Food samples

Test Marketing
Controlled experiment conducted on a small segment of the target market Major objectives
Determine how well products will be accepted in the marketplace Determine how changes in marketing mix will likely affect product success

Major reason for test marketing is risk reduction

Lose $ 1 million in test market or $ 50 million on product failure?

Problems
Expense Time Competitors can disrupt

Factors to Consider
Population size Demographic composition Lifestyle considerations Competitive situation Media coverage & efficiency Media isolation Self-contained trading area Overused test markets Loss of secrecy