CHAPTER 8 CAUSAL RESEARCH DESIGN: EXPERIMENTATION OPENING QUESTIONS 1. 2. 3. 4. 5. 6. 7. 8. 9. How is the concept of causality defined in marketing research and how do we distinguish between the ordinary meaning and the scientific meaning of causality? What are the conditions for causality? Can a causal relationship be demonstrated conclusively? How do we define and differentiate the two types of validity: internal and external? What are the various types of experimental designs, and what are the differences among preexperimental, true experimental, quasi-experimental, and statistical designs? How can we compare and contrast the use of laboratory versus field experimentation and experimental versus nonexperimental designs in marketing research? What is test marketing and how does it involve experimentation? Why is the internal and external validity of field experiments that are conducted overseas generally lower than in the United States? How does technology facilitate causal research? What ethical issues are involved in conducting causal research, and how can debriefing address some of these issues? AUTHOR'S NOTES: CHAPTER FOCUS This chapter provides an overview of causal research and experimentation. The concept of causality and the conditions for causality are described. Internal and external validity in experimentation are discussed in detail. A classification of experimental designs is presented and pre-experimental, true experimental, quasi-experimental, and statistical designs are described. Comparisons between laboratory versus field experiments, and experimental versus nonexperimental designs are made. Test marketing is presented as an application of experimental designs. This chapter covers the material on causal research contained in competing texts; yet, the treatment of experimentation is more extensive than that found in competing texts. This chapter could be taught by focusing on the opening questions sequentially. Greater emphasis could be placed on the concept and conditions for causality (Opening Questions 1 and 2), internal and external validity (Opening Question3), and the various types of experimental designs (Opening Question 4). However, quasi experimental designs and statistical designs may not be covered in detail. Conducting experiments internationally (Opening Question 7), ethical issues (Opening Question 9), and use of the Internet should be discussed. CHAPTER OUTLINE 1. 2. 3. Overview Concept of Causality Conditions for Causality i. Concomitant Variation ii. Time Order of Occurrence of Variables 103 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. iii. Absence of Other Possible Causal Factors iv. Role of Evidence What is Experimentation? Definition of Symbols Validity in Experimentation i. Internal Validity ii. External Validity A Classification of Experimental Designs Pre-experimental Designs i. One-Shot Case Study ii. One-Group Pretest-Posttest Design iii. Static Group True Experimental Designs i. Pretest-Posttest Control Group Design ii. Posttest-Only Control Group Design Quasi Experimental Designs i. Time Series Design Statistical Designs i. Factorial Design Selecting an Experimental Design Laboratory Versus Field Experiments Limitations of Experimentation i. Time ii. Cost iii. Administration Application: Test Marketing Summary Illustration Using the Opening Vignette International Marketing Research Technology and Marketing Research Ethics in Marketing Research Summary Key Terms and Concepts Acronyms TEACHING SUGGESTIONS Opening Question 1 * Discuss the concept of causality as used in marketing. You may want to begin the class by asking a few students what the term ‘causality’ means and writing components of their definitions on the board. Afterwards, be sure to stress to the students that marketing effects are caused by multiple variables and the relationships tend to be probabilistic. As such, it is not possible to conclusively prove causality. Thus, we can only infer a cause-and-effect relationship between the variables. * Via an example, explain the three necessary conditions to infer causality - 104 1. 2. 3. Concomitant variation: is the extent to which a cause, X, and an effect, Y, occur together or vary together in the way predicted by the hypothesis under consideration. Evidence pertaining to concomitant variation can be obtained in a qualitative or quantitative manner. Thus, we may hypothesize that increasing product price will lead to an increase in profits. Time order of occurrence of variables: the causing event must occur either before or simultaneously with the effect; it cannot occur afterwards. Thus, our profits cannot increase until after we increase the price. Absence of other possible causal factors: means that we can infer causality if all other factors affecting the subject are absent, i.e., these factors are held constant. Thus, other factors like store penetration, expanded consumer spending, or decreased costs must be held constant. * Review the key definitions related to experimentation - 1. 2. 3. 4. 5. 6. Independent variable: variables or alternatives that are manipulated and whose effects are measured and compared, such as price levels, different package designs, and advertising themes. Test units: individuals, organizations, or other entities whose response to the independent variables or treatments is being examined, such as consumers, stores, and geographic areas. Dependent variables: variables that measure the effect of the independent variables on the test units, such as sales, profits, and market shares. Extraneous variables: all the variables other than the independent variables or treatments that affect the response of the test units to the treatments, such as store size, store location, and competitive effort. Experiment: the process of manipulating one or more independent variables to determine their effect on the dependent variable, controlling for the effect of extraneous factors. Experimental design: a set of procedures specifying (1) the test units and how these units are to be divided into homogeneous sub-samples; (2) independent variables or treatments that are to be manipulated; (3) dependent variables are to be measured; and (4) how the extraneous variables are to be dealt with. * Introduce the symbolic notation used in marketing research. 1. X = 2. O = 3. R = the exposure of a group to an independent variable or treatment or event, the effects of which are to be determined the process of observation or measurement of the dependent variable on the test units or group of units the random assignment of test units or groups to separate treatments The following conventions are associated with the symbols: 1. Movement from left to right indicates movement through time. 2. Horizontal alignment of symbols implies that all those symbols refer to a specific treatment group. 105 3. Vertical alignment of symbols implies that those symbols refer to activities or events that occur simultaneously. Thus, the first example in the text, X O1 O2, indicates a treatment was given to a sample and measurements at two different points in time were made. Opening Question 2 * Discuss the conditions for causality. At least three conditions must be satisfied in order to justify the inference of a causal relationship between two variables: (1) concomitant variation, (2) time order of occurrence of variables, and (3) absence of other possible causal factors. These conditions are necessary but not sufficient to demonstrate causality. By that we mean that they must be satisfied to justify drawing a causal inference; however, their presence does not guarantee that we have isolated the true variables responsible for the effects we are observing. Describe and illustrate the three conditions to the students. Opening Question 3 * Define validity and describe the difference between internal and external validity. Start by defining validity, in an experimental setting, as the ability to draw appropriate conclusions about the effects of independent variables and to make reasonable generalizations to a larger population of interest. With this definition, we can distinguish internal and external validity. Internal validity examines whether the manipulation of the independent variables or treatments actually caused the effects on the dependent variables, while external validity examines whether the cause-and-effect relationships found in the experiment can be generalized. Because of the conceptual difficulty in understanding validity, some examples that depict a research design lacking one of the two types of validity would be instructive. For example, if we hypothesize that selling through national chains will increase profits, we must hold all other causes of increased profits constant in order to conclude internal validity. External validity can be seen in a study that randomly selects stores from all markets in which a national chain operates. Assuming that the study takes into account all realistic factors, any findings that hold for all these markets will generalize to all stores in the national chain. Opening Question 4 * Discuss the differences in the types of experimental designs. 1. 2. 3. Pre-experimental Designs: do not apply randomization, thus, they can be influenced by extraneous factors. True Experimental Designs: randomization is included, thus, allowing researchers to randomly assign test units and treatments to experimental groups. Quasi Experimental Designs: lacks control over the scheduling of treatments and the ability to randomly expose test units to treatments. 106 4. Statistical Designs: a series of several basic experiments that are conducted simultaneously. These designs suffer from the same shortcomings as the actual designs of the basic experiments. Figure 8.4 can be used here to organize your discussion on experimental designs. * Describe pre-experimental designs and their shortcomings - Note that pre-experimental designs include the one-shot case study, the one-group pretestposttest design, and the static group design. Stress that the salient feature of these designs is that they do not employ randomization procedures to control for extraneous factors. For example, in the one-group pretest-posttest design, an observation on attitudes toward the brand is taken before and after watching a commercial, but randomization is not used in selecting observers; thus, the results cannot be extended to the population as a whole. * List the significant features of a true experimental design and provide examples. First, to introduce the ensuing discussion, distinguish true experimental designs by the fact that the researcher can randomly assign test units to experimental groups and also randomly assign treatments to experimental groups. Examples of True Experimental Designs: 1. Pretest-Posttest Control Group Design: this design controls for most of the extraneous variables. Selection bias is eliminated by randomization, and interactive testing effect is not controlled due to the effect of the pretest measurement on the reaction of units in the experimental group to the treatment. 2. Posttest-Only Control Group Design: compared to the Solomon four-group design, this design is fairly simple to implement because there is no premeasurement, thus, the testing effects are eliminated. However, this design is sensitive to selection bias and mortality. 3. Solomon Four-Group Design: the Solomon four-group design overcomes the limitations of the pretest-posttest control group and the posttest only control group designs in that it explicitly controls for interactive testing effect, in addition to controlling for all extraneous variables. This design allows the researcher to update and measure the various effects. It will prove instructive to use a running example in explaining these designs. A standard example is to use advertising research in the measurement of consumer attitudes toward a commercial or the brand being advertised. * Explain the analysis of quasi experimental designs. Mention that in a quasi experimental design, the researcher can control when measurements are taken and on whom they are taken, but the researcher is lacking control over the scheduling of the treatments and the ability to randomly expose test units to the treatments. Quasi experimental designs are quicker and less expensive than true experimentation, but should only be used when a true experimental design is not feasible. 107 * Explain the nature and uses of statistical designs. Describe statistical designs as a series of basic experiments that enable statistical control and analysis of external variables; thus several basic experiments are conducted simultaneously. Advantages: 1. The effects of more than one independent variable can be measured. 2. Specific extraneous variables can be statistically controlled. 3. Economical designs can be formulated when each test unit is measured more than once. Examples of a randomized block design, Latin Square design, and Factorial design can be constructed for demonstration purposes. See Table 8.1 Opening Question 5 * Start a class discussion on whether laboratory or field experiments are more useful in marketing research. Begin this discussion by simply asking students which form of experimentation is better. You may want to note that laboratory experiments make up the bulk of consumer research because of their ability to control extraneous variables and their relative efficiency in gathering data. However, for certain studies, field experiments are used. For example, Coca-Cola counts the shelf space it and its competitors receive in local grocery stores when promotional variables are manipulated. Thus, each type of experimentation has its role to play in marketing research, but because causality cannot be inferred from field experiments, laboratory experiments predominate. See Table 8.2 for a summary of laboratory vs. field experimentation. * Discuss why nonexperimental designs are limited for inferring cause-effect relationships. Mention that descriptive studies are sometimes used to infer causality, but this may not be appropriate. In descriptive studies, prior equivalence of the groups with respect to both the independent and dependent variables is difficult to establish, as is time order of occurrence of the variables. Finally, it offers little control in eliminating other extraneous factors. * Discuss the limitations of experimentation Start by asserting that theoretically, experimentation is a very powerful way to uncover causal relationships. However, in practice, real world considerations prevent it from achieving its optimal effectiveness. These considerations are: 108 1. 2. 3. Time: experiments can be time consuming; however, the experiments should be long enough in duration so that the post-treatment measurements include most, or all of the effects of the variables. Cost: the requirements of experimental group, control group, and multiple measurements significantly add to the cost of the experiment. Administration: experiments can be difficult to administer, and it is not always feasible to control for the effects of the extraneous variables. Opening Question 6 * Define and classify test marketing. Define test marketing as an application of a controlled experiment, done in a limited but carefully selected part of the marketplace called test markets, which involves a replication of the planned national marketing program for a product. In a standard test market the product is sold through regular distribution channels. The company's own salesforce is responsible for distributing the product, stocking the shelves, and taking inventory at regular intervals. A controlled test marketing program is conducted by an outside research company. The research company is responsible for all activities associated with the test market. Simulated test markets yield mathematical estimates of market share based on initial reaction of the consumers to the new product. Opening Question 7 * Examine the difficulties involved with international experiments. It is difficult to control for the time order of occurrence of variables and it is difficult to control for the absence of other possible causal factors, two of the necessary conditions for causality in an experiment. • Control of extraneous variables is particularly problematic. • It may not be possible to address this problem by adopting the most appropriate experimental design, as environmental constraints may make that design infeasible. • The internal and external validity of field experiments conducted overseas is generally lower than in the United States. EXAMPLE: M&M/Mars has set up massive manufacturing facilities in Russia and advertises its candy bars on television. Yet, sales have been low. Is Mars advertising too much, too little, or just right? While the answer could be determined by conducting a field experiment that manipulated the level of television advertising, such causal research is not feasible given the tight control of television programming and advertising by the Russian government. TV stations are owned and operated by the government with severe restrictions on television advertising. This makes field experiments manipulating advertising levels extremely difficult. 109 Opening Question 8 * Describe how technology is facilitating experimentation. Virtual Reality (VR) is a real-time, 3D environment made to represent either reality or an environment out of someone's imagination. These environments are created by high-powered computer systems. As with many of the technological innovations used in marketing research, VR was not developed specifically with research in mind, but VR is finding its way into the field of marketing research nonetheless and holds great promise for conducting causal research. Using VR, the researcher can create an environment that represents the field (marketplace) and yet exercise the degree of control possible only in a laboratory setting. Discuss Visionary Shopper by MarketWare Simulation Services. Opening Question 9 * Discuss the ethical responsibilities resulting from disguised experiments. It is often necessary in experimentation to disguise the purpose of the research to produce valid results. Disguising the purpose of the research should not lead to deception, however. Although this seems like a paradox, one solution would be to disclose the possible existence of deception before the start of the experiment and allow the participants the right to redress at the conclusion of the experiment. The following four items should be conveyed: (1) inform respondents that in an experiment of this nature a disguise of the purpose is often required for valid results, (2) inform them of the general nature of the experiment and what they will be asked to do, (3) make sure they know that they can leave the experiment at any time, and (4) inform them that the study will be fully explained after the data have been gathered and at that time they may request that their information be withdrawn. The procedure outlined in item (4) is called debriefing. One further ethical concern in experimentation involves using the appropriate experimental design to control errors caused by extraneous variables. It is the responsibility of the researcher to use the most applicable experimental design for the problem. Determining the most appropriate experimental design for the problem requires not only an initial evaluation but also continuous monitoring. Internet Emphasis * Indicate that the Internet and computers are helpful in the analysis of experimental design data. Different experimental treatments can be displayed at different Web sites. Respondents can then be recruited to visit these sites and respond to a questionnaire that obtains information on the dependent and extraneous variables. Thus, the Internet can provide a mechanism for controlled experimentation, although in a laboratory type of environment. To complement the Internet, microcomputers and mainframe software can be used in the design and analysis of experiments. The comprehensive statistical-analysis software package Minitab can be used to design experiments. Although similar in use to, SPSS or SAS, Minitab 110 includes functions and documentation specifically for industrial quality control work where factorial designs are encountered. BE AN MR! AND BE A DM! It should be noted that a variety of answers are appropriate. The ones given here are merely illustrative. Be a DM!: FTC: To ensure the smooth operation of our free market system, the Federal Trade Commission (FTC) enforces federal consumer protection laws that prevent fraud, deception and unfair business practices. The Commission also enforces federal antitrust laws that prohibit anticompetitive mergers and other business practices that restrict competition and harm consumers. Whether combating telemarketing fraud, Internet scams or price-fixing schemes, the FTC’s primary mission is to protect consumers. - Information overflow; difficult for customers to differentiate noise and useful information - Information flow cannot be controlled or traced. This can lead to misrepresentation of information, fraud etc. - Dumping of products especially from foreign countries (web based selling of drugs by Canadian and Indian companies) - Increased possibility of collusion between companies Be an MR!: FTC: The conditions necessary to conclude such causality are: 1. Concomitant variation - the information available has increased with increased consumer use. 2. Time Order - increased information lags or concurs with increased information use. 3. Absence of other possible causes Be an MR!: Fox News Information on consumers’ preferences for network news channels can be obtained from sources such as http://www.rateitall.com/t-8-tv_news_shows.aspx. A preexperimental design would not be appropriate as there are three experimental groups. A statistical design is called for. Be a DM!: Fox News The popularity ratings for channels can be used to assess the attributes that audiences look for in Network News Channels. Based on experimentation, causality can be assigned to factors that make the news more attractive. Improve the attributes preferred by audience. Convey the change in form and content to audience through appropriate PR and Advertising channels. EXPERIENTIAL LEARNING Canon 1. Students should be encouraged to visit www.bestbuys.com and identify the relevant information. 2. Students should be encouraged to search the Internet and identify the relevant information on consumers’ price sensitivity for digital cameras. 111 3. 4. A factorial design is appropriate. The price of digital cameras could be varied and consumers’ willingness to buy measured. It would be important to control for income and digital camera usage. It is likely that the market is segmented with respect to price sensitivity. Different prices can be charged to the different segments with some differentiation in the cameras offered. A Local Pizza Maker 1. Students should be encouraged to search the Internet and identify the relevant information. 2. It would be highly beneficial for the firm to conduct a factorial experimental sign because this involves measuring the effects of two or more independent variables at various levels. In addition, it is highly recommended that management test market in their largest store in the local area. Here is our proposed format: a. Controlled Variables: While test marketing at a store location, the amount of pizza sauce will remain unchanged as respondents determine if their preference is a 2:1 ratio of provolone/mozzarella or a 1:2 ratio. Another ingredient, such as crust, will also remain the same so that it would prevent undue influence upon each respondent’s perceptions of the new pizza formula. Store location will also be non-factor so that respondents do not disproportionately influence their perceptions on price and pizza formulation. b. Manipulated Variables: One such manipulated independent variable could include evaluating various promotional tactics, versus none at all, when evaluating price and the new provolone/cheese mix concept. For example, the team could evaluate such measures like “Buy One, Get One Free” promotions at the local pizza maker’s store as part of the test market strategy. Management would also determine how this particular manipulated independent variable affects ratings for the new pizza formulation. When evaluating the new pizza cheese mix, it would be imperative to gauge the price level at three stages (raised, lowered or unchanged) to one that is kept constant and its affect on ratings. It would also be informative to manipulate the ratio of pizza cheese (2:1 or 1:2) and evaluate the consumer’s response level. c. Response Variables: One obvious dependent variable that would be measured is consumer’s perceptions of overall brand quality ratings. Other measures that would be useful are consumers overall preference, and willingness to buy. d. Control Group: One experimental treatment will consist of the current price and the current cheese mix; this can serve as a control group. A separate control group is not required. 3. Although costs and time allocation is of significant important, a field experiment would be more useful because our client is seeking external validity in which a cause-effect relationship found in the experiment could be generalized. It also provides a realistic forum for our experimental design in which our results could benefit our client more effectively. Most importantly, it enables the researchers to evaluate the behavior, attitudes and perceptions of pizza customers in a natural environment. We suggest a test market for the new provolone/mozzarella cheese mix in the largest store in the Metroplex or largest shelf-space at a local convenient store. This would determine if the new pizza product should expand to all locations for full market rollout. 112 4. Incentives for this project could total $3600 dollars. 180 respondents x $20 (for participation)=$3600 (see Table 1 below) Table 1. Factorial Design Provolone/Mozzarella Raise Price Constant Price Lowered Price 2:1 ratio 30 respondents 30 respondents 30 respondents 1:2 ratio 30 respondents 30 respondents 30 respondents Total Incentives $3600 5. I would probably switch to the less time consuming and fiscally conservative method of a laboratory environment. Although the research team would have more control of the environment, it is more subject to reactive error, perhaps low in external validity and is more artificial when evaluating consumer behavior. 6. Pizza Hut can increase its market share by using the experimental results to identify the most desired combination of cheese formulation and price. WHAT WOULD YOU DO? It should be noted that a variety of answers are appropriate. The ones given here are merely illustrative. The Marketing Research Decision 1. A, B & E 2. A lot of secondary data about apparel industry is available from online as well as traditional third party databases. This can be used initially to gather information about the success of private labels and in-house brands. The focus groups are helpful in gaining further understanding of the conclusions drawn from the analysis of secondary data, understanding consumer priorities, perceptions and preferences. The specific causal design that is recommended is a factorial design. The factor, brand name can be varied at three levels: Levi’s, Gap, and Arizona. Respondents can be randomly assigned to one of these brands and asked to express preference for the brand rate its power. Current usage of these brands should be controlled. 3. The causal design, as specified, will provide a direct relative measure of the power of the three brands when the mean levels are compared. 113 The Marketing Management Decision 1. B and C 2. Comparing the prices, it appears to be one of the reasons why Levi’s is loosing the battle for market share. It would be a good idea to launch a cheaper line of jeans to appeal to the cost conscious market segment. To reach the new target segment, the company would want to increase advertising spend in targeted marketing efforts. The new line of jeans should be distributed through different channels than that of the existing high priced segment. This will ensure the necessary reach. REVIEW QUESTIONS 1. 2. 3. 4. There are three requirements or conditions that must be satisfied before a causal relationship can be inferred. The three conditions are: 1. Concomitant variation of the two variables 2. Time order of occurrence of variables 3. Elimination of other possible causal factors The difference between internal and external validity is that internal validity is concerned with the question of whether the observed effects on the dependent variable have been caused by variables other than the treatment, and control of extraneous variables is necessary for establishing internal validity. External validity, on the other hand, is concerned with whether the cause-effect relationship can be generalized. More often than not a tradeoff is required between internal and external validity. While internal validity can be raised in an artificial environment, this reduces the generalizability of the results and consequently, the external validity. The distinguishing feature of the true experimental design, as compared to the preexperimental design, is randomization. In true experimental designs, the researcher can randomly assign test units to experimental groups and also randomly assign treatments to experimental groups. The steps involved in implementing the posttest-only control group design are: 1. Select a sample of respondents at random. 2. Randomly assign the respondents to two groups (i.e. experimental group and control group). 3. Respondents in the experimental group would then be exposed to a treatment. 4. Post-treatment measurements are obtained from both groups using a measuring instrument like a questionnaire. The design is described symbolically as follows: EG: R X 01 CG: R 02 5. Treatment effect is given by: TE = 01 – 02 A time series experiment is a quasi experimental design. It involves periodic measurement on the dependent variable for a group of test units, and then the treatment is administered by the researcher or occurs naturally. After the treatment, periodic 114 measurements are continued in order to determine the treatment effect. A time series experiment may be symbolically described as: 01 02 03 04 05 X 06 07 08 09 010 6. 7. 8. 9. This design is used in cases where the researcher lacks control over the scheduling of the treatment and the ability to randomly expose test units to the treatment, but can control when measurements are taken and on whom they are taken. A multiple time series design is different from the basic time series in that a multiple time series design has an additional group of test units added to serve as a control group. The advantages statistical designs have over basic designs are: a. In the case of basic designs, the effects of only one independent variable or treatment can be measured, while in case of a statistical design, the effects of more than one independent variable can be measured. b. In the case of statistical designs, specific extraneous variables can be statistically controlled. c. Economical designs can be formulated when each test unit is measured more than once. A laboratory experiment is one where the researcher creates an artificial environment for the purpose of experiment; thus, he has a great degree of control over extraneous variables. On the other hand, a field environment is one where the experiment is conducted under actual conditions. Laboratory experiments have some advantages and some disadvantages vis-à-vis field experiments. The advantages are that there is a greater control over extraneous variables, internal validity is high, and time required is short. It is also economical. The disadvantages are that the environment is artificial and external validity is low. Therefore, the results cannot be safely generalized. Also, the reactive error is high due to artificiality of the environment, which may cause the respondent to react to the situation itself, rather than to the independent variable. Test marketing is an application of a controlled experiment done in a limited but carefully selected part of the marketplace called test markets. It involves a replication of a planned national marketing program for a product in the test markets. The three types of test markets are: 1. Standard test market: The product is sold through regular distribution channels and, typically, the company's own salesforce is responsible for distributing the product. It involves a one-shot case study. 2. Controlled test market: The entire test marketing program is conducted by an outside research company, which also handles distribution and field sales operations in the test market. 3. Simulated test market: These are also called laboratory tests and yield mathematical estimates of market share based on initial reaction of consumers to the new product. 115 APPLIED PROBLEMS 1. 2. 3. 4. 5. 6. 7. 8. a. Independent Variable - The commercial Dependent Variable - Attitude towards abortion b. Posttest-only control group design External validity may be affected because this was a lab experiment, not an actual field experiment, thus, the results may not be generalizable to those obtained when the commercial is actually viewed in home. If the respondents had been selected by convenience, a static group design would result. Any experimental effect could also be attributed to selection or mortality. a. This is an example of the one-shot case study. The internal validity is threatened by history, maturation, selection, and mortality. b. This is an example of the posttest-only control group design. Internal validity is not hampered by extraneous variables. Many answers are possible for each design. One solution is provided as an example. a. Measuring the attitude towards a new advertisement or new product b. Evaluating the effectiveness of a change in a product attribute or the effectiveness of a new marketing program when time and cost are not major considerations c. Evaluating the attitude toward a new advertisement or product when time and cost are not significant factors, forcing researchers to reduce sample size, quickly determine results, and/or keep costs down d. When great precision is needed and the effectiveness of an advertising program needs to be monitored over time to show the process of change in the population e. A new marketing program is being initiated which must be evaluated to determine its effectiveness. The goal is to determine which of the components of the program (price, promotion, channels, etc.) are the most important factors leading to sales and to assess the interaction of these components on sales. Several of the designs can be implemented. A two factor design, where factor one denotes the value of coupon (none, low, medium, and high), and factor two denotes product use (none, low, medium, and high), seems to be very appropriate. Students should select a specific product and also define the levels (low, medium, and high) for each factor. The existing and the three new packages can be displayed, one at each specific location on the Internet. Respondents can be recruited and randomly assigned to each package location (URL), asked to view the package and answer several attitudinal and behavioral intent questions. The package with the most favorable attitude and/or the highest behavioral intent can then be identified. The previous (existing) and the new version of EXCEL can be made available, one at each specific location on the Internet. Respondents can be recruited and randomly assigned to each EXCEL version, and then asked to do a specific task using that version. Then they answer several questions relating to ease of use, performance, features, and overall preference. The most favorable version can then be identified. Respondents would be randomly recruited and randomly assigned to one of two Internet locations (URLs). The experimental group of respondents would first see the print ad for Toyota Camry and then respond to a battery of attitudinal and behavioral intent questions. The control group respondents would not see the print ad but respond to the same battery 116 of attitudinal and behavioral intent questions. The effectiveness of the print ad can be determined by comparing the responses of the two groups. GROUP DISCUSSIONS 1. Issues which can be discussed include the role of inference in drawing conclusions, the difficulty of isolating all possible sources of variation in social science experimentation, the historic value of experimentation in deriving theories and inferring relationships, and the nature or definition of science, (i.e., is science only limited to proofs which account for all sources of variation?). 117