Chapter 5: Descriptive and Causal Research

Key Differences (Experimentation vs. Causal)

1. In an experiment, one or more of the independent variables are deliberately manipulated while others are controlled. 2. Combinations of conditions are assigned to sample elements on a random basis. This reduces the likelihood of preexisting conditions affecting results.

The Use of Test Markets

1. It is necessary to weigh the cost and risk of product failure against the profit and profitability of success 2. The greater the investment involved, the more valuable a test market becomes 3. Likelihood and speed with which the competition will be able to copy the product

Describe Longitudinal Studies

1. Primary objective is to monitor behavior over time and thereby identify behavioral changes 2. Referred as true panels: they report the same information at specific points in time

Test Market Designs

1. Standard test market uses existing marketing channels. 2. Controlled store test market use stores selected for the test. 3. Simulated test markets uses respondent feedback in products as test data.

Experimental Design Symbols

1. X = exposure of a group of subjects to an experimental treatment or a level of an independent variable. If different levels are used, then X1, X2,...Xn are used. 2. O = observation or measurement of the dependent variable in which the researcher is interested. 3. R = random assignment of people to groups 4. Left-to-right notations indicates the time sequence of occurrence of events. 5. All the notations on a single line refer to a single group of respondents, and notations that appear together vertically identify events that occurred at that same point in time.

External Validity

1.External validity: is concerned with whether the results are generalizable to other subjects, stores, or areas. Major Threats: Subject Selection, Subject Response Subject selection: the population from which subjects are selected may not be the same as the target market. Other factors: the experimental settings themselves may have an effect on a subject's response. If the subjects know they are participating in a price experiment, they be more sensitive to price than normal.

Field vs. Laboratory Experiments

1.Laboratory experiments: are experiments conducted under artificial conditions, such as testing television ads in a movie theater rather than the buyers homes 2.Field experiments: result in an experimental setting which is more realistic in terms of modeling actual conditions and is therefore higher on external validity. They are carried out in a natural setting with a minimum of artificial elements in the experiment.

3 Things Causal Research Does:

1.Manipulation: we manipulate the causal or independent variable such as ad message or price. 2.Measure: we measure the effect or dependent variable such as ad recall or profit. 3.Control: we control other variables which could have an impact on our dependent variable.

Characteristics of Cross-Sectional Designs

1.Most frequently used 2.Involves a sampling of a population of interest at one point in time 3.Referred to as sample surveys 4.Omnibus panels: which consists of hundreds of thousands of U.S. households that have been selected to represent the U.S. population.

Descriptive Research Objectives

1.To describe the characteristics of relevant groups (such as the 20 percent of our customers who generate 80 percent of our business.) 2.To determine the extent to which two or more variables covary. 3.To estimate the proportion of a population who act a certain way. 4.To make specific predictions.

Control Group

Control group: is a randomly selected group of units in the experiment. It does not receive a treatment and thus serves as a benchmark for comparing the effects of treatment.

Types of Descriptive Designs

Cross-Sectional and Longitudinal

Pre-experimental designs

Designs that are weak in terms of their ability to control the various threats of internal validity. One-shot case study X1O1 The one-group pretest-posttest O1X1O2 The static-group comparison x1O1 O2

Experimental Design

Experimental design: is the specific process used to arrange independent variables into treatment and then assign treatment to units. They create the variables for treatments. Then, assign treatments/variables to units.

Experimental vs. Casual Design

Experimental or casual designs differ from other research designs in the degree of control exerted by the researcher over the conditions under which data are collected.

4 Types of Experimentation

Experimental treatments, Experimental units, Experimental Design, and Control Group

Experimental Treatments:

Experimental treatments: a specified combination of independent variables. Testing different methods on a same product. E.g. if an experiment was designed to measure the influence of three different levels of advertising on sales, then each level of advertising would be a different treatment.

Experimental Units

Experimental units: are the geographic areas, stores, or people whose responses are measured in determining the effect of the different treatments

Limitations of Causal Research

Field experiments can involve many variables outside the control of the experimenters, resulting in unanticipated differences in conditions surrounding treatment groups. It may be difficult or expensive to gain the cooperation of retailers. Marketing personnel may lack knowledge of experimental procedures, reducing the chance of results which demonstrate causality. Experiments are notoriously expensive and time consuming. The experiment must be careful not to introduce bias into the experiment by saying or doing something that may consciously affect the behavior of the test participants.

Two Types of Validity

Internal and External Validity

Internal Validity (Definition and Threats)

Internal validity: has to do with whether the independent variables that were manipulated caused the changes in the dependent variable. 7 Major Threats; History, Maturation, Testing, Instrumentation, Selection, Statistical Regression, and Mortality History: during the time that an experiment is taking place, some events may occur that affect the relationship being studied. Maturation: changes may take place within the subject that are a function of a passage of time and are not specific to any particular event. Testing: when pretreatment and postreatment measures are used, the process of taking a test can affect the scores of a second measures. Instrumentation: this comes from changes in measuring instruments or observers. Using different questions or different observers or interviews is a validity threat. Selection: one of the most important threats is the selection of subjects to be included in experimental and control groups. The concern is over initial differences that exist between subjects. Statistical regression: is of special concern when subjects have been selected on the basis of their extreme scores. Mortality: occurs when the composition of the study groups changes during the experiment.

quasi-experiment

Nonequivalent control group: most widely used. Groups are not randomly assigned. O1X1O2 O3 O4 Separate sample pretest-posttest: used in situations that cannot control when and to whom to introduce the treatment but can control when and whom to measure. R O1 RX1O2

Test Marketing Uses

Potential uses: 1. The sales volume and market share expectations of new product 2. Some estimate of the repurchase cycle and likelihood of repurchase 3. A profile of trier-adopters 4. An understanding of competitors reactions to a new product 5. Some feel for the effect of the new product sales on existing product sales 6. Performance of new product package design in generating trial and satisfaction

3 Experimental Research Designs

Pre-Experiments, True Experiments, and Quasi-Experiments

true experiemental designs

Pretest-posttest control group RO1X1O2 RO3 O4 Posttest-only control group RX1O1 R O2

Causal Research

When we are in need of determining if two or more variables are causally related we must turn to causal research procedures. Our interest in causal research is to determine the degree to which one variable is causally related to another

Ex post Facto Research

try to discover a causal relationship without manipulation of the independent variable or control over respondent exposure to treatment conditions. It is followed by attempts to find the causal factor that caused the observed outcome.