Multivariate analysis of parametric data
For CASE5, you are to examine several articles and come to some conclusions regarding the appropriateness of using ANOVA vs. other GLM techniques.
CASE5.1 - GLM (General Linear Model)
Maddox, Lynda M (1999). The use of pharmaceutical Web sites for prescription drug information and product requests. Journal of Product and Brand Management, 8(6). 488-496.
If the above link does not work, note that Maddox (1999) was available on 2012 January 30 in Emerald-Library via the Trident Univeristy / Touro College eLibrary.
Lynda M Maddox (1999) asked her respondents a series of questions about why they visited a pharmaceutical Web site. These reasons included learning about a particular disease or drug, deciding which drug is right, getting a second opinion on a diagnosis, getting anonymous information from a medical expert, and being able to discuss a subject more knowledgeably with their doctor.
Responses to the survey questions were rated on a five-point scale ranging from 1 (very unimportant) to 5 (very important).
Respondents were next asked questions about their behavioral responses regarding how likely they were either to request more information about a drug from their doctor or to ask the physician to prescribe a particular medication.
Responses to the survey questions were rated on a five-point scale ranging from 1 (very unlikely) to 5 (very likely).
Age and gender were also recorded.
In terms of the results:
Maddox (1999) said that there were 132 usable questionnaires. These usable questionnaires were completed by subjects, respondents, or cases. So, there were 132 usable cases. So, in the lingo of research, several pieces of data were collected from and on each case.
The primary reason for visiting the Web site was to learn about a particular medicine or drug, with 55 percent rating this reason as important or very important. This was followed by learning about a particular disease (48 percent important or very important), getting a second opinion on a medical diagnosis (43 percent important or very important), getting anonymous information from a medical expert (34 percent important or very important), deciding which medication or drug is right for them (34 percent important or very important), and, last, to be able to discuss knowledgeably with their doctor (25 percent).
A total of 36 percent said they would be likely or very likely to request more information from their doctors. A further 32 percent said they would actually ask their doctor to prescribe a particular medication. In Maddox (1999), the means reported in Table I reflect these data.
On average, females gave higher importance and higher likelihood ratings than males on all questions. A total of 70 percent of women versus 42 percent of men said they went to the Web site to learn more about a particular medicine or drug, and 45 percent of women and 24 percent of men went to decide which drug was right for them. A total of 69 percent of women and 32 percent of men wanted to learn more about a particular disease. Further, 57 percent of women and 32 percent of men wanted to get a second opinion on a medical diagnosis, and 43 and 24 percent respectively wanted anonymous information from a medical expert.
A total of 45 percent of women and 30 percent of men were likely or very likely to request more information from their doctor on a particular product; while 37 percent of women and 27 percent of men were likely or very likely to ask their doctor to prescribe a particular medication.
Results
T-tests were used to analyze the data. All significance testing was at the 0.05 significance level (p < 0.05).
Gender differences
On average, females gave higher importance and higher likelihood ratings than males on all questions.
In matched-sample t-tests conducted separately on each rating, women were significantly higher than males in every case.
There were no significant interactions between gender and the reason that was rated, however, suggesting that the difference between men and women was statistically the same for all the reasons rated.
Age differences
As is shown in Table II in Maddox (1999), on average, respondents aged 50+ gave higher importance ratings than did the younger respondents. For three of the importance ratings (get a second opinion, decide which drug is right, and discuss subject knowledgeably with doctor), the means for respondents aged 50+ were significantly higher than the means for at least one of the younger age groups, according to t-tests
Please answer the following questions:
5.1. What is Maddox's research design?
5.1a. Do you agree with the following statement?
Maddox (1999) should have used a Repeated Meassure ANOVA to test for differences (based on AGE and GENDER) in the set of responses.
In answering this and the following questions, remember that a quasi-experiment (eg demographics as independent variables) can be analyzed using the techniques of experimental design. So, even if the people were not assigned to groups, you can still analyze the differences between the groups as if the people were randomly assigned to those groups.
Defend your answer.
5.1b. Do you agree with the following statement?
Maddox (1999) should have used a Multipoe Factor MANOVA to test for differences (based on AGE and GENDER) in the set of responses.
Defend your answer.
5.1c. Do you agree with the following statement?
To show that the effect of AGE and GENDER on each of the dependent variables holding constant the effects of co-variates, Maddox (1999) should have used a Repeated Measure ANCOVA.
Defend your answer.
5.1d. Do you agree with the following statement?
To show that the effect of AGE and GENDER on each of the dependent variables holding constant the effects of co-variates, Maddox (1999) should have used a Multiple Factor ANCOVA (that is, Maddos (199) should have used a MANCOVA).
Defend your answer.
5.1e. Do you agree with the following statement?
Maddox (1999) should have examined the responses to each of her questions using Multiple Regression, using 0 and 1 for Gender and 1, 2, 3, and 4 for AGE, and should have computed partial correlations thereby holding the effects of one of the predictor variables constant while she examined the effect of the other predictor variable.
Defend your answer.
CASE5.2 - MANOVA (Multivariate Analysis of Variance)
Bello, Daniel & Williamson, Nicholas (1985, Fall). The American Export Trading Company: Designing a New International Marketing Institution. Journal of Marketing, 49(4). 60-69.
Available on 2011 January 19 in EBSCOhost via the Trident Univeristy / Touro College eLibrary .
Daniel Bello & Nicholas Williamson (1985) use MANOVA and ANOVA to test four hypotheses.
In hypothesis 1, Bello & Williamson (1985) state that:
The importance of services provided by export intermediaries is influenced by the type of product exported, export role of the intermediary, and supplier's export sales volume.
In hypothesis 2, Bello & Williamson (1985) state that:
Type of product influences the importance of services provided by export intermediaries in the following manner:
Transaction creating services are more important for differentiating products
Physical fulfilling services are more important for undifferentiated products.
Please do the following tasks:
5.2 Describe the research design.
5.2a. There are so many variables described in this article that it would be helpful to me, if you would tell me who (or what) are the cases.* Then, using the format for a codebook, list the variables (add as many rows as you need to add to the following table), and list value labels for each of those variables.
- Case is a term that was coined by the writer's of BMP, a computer program that was the precursor to BMPD, SAS. SPSS, etc. It was chosen by those writers because BMP was originally written for BioMed research where, in the medical sense, a Cases are individuals. Today, at least in business lingo, a case can be an individual, a division of a company, a company, an industry, a state, a region, a country, or whatever the researcher deems appropriate.
Variables Value Labels
1.
2.
...
n.
1.
2.
...
n.
5.2b1. Choose either 5.2b1 OR 5.2b2.
5.2b1a. Based on Hypothesis 1, write out a multiple regression equation using as your dependent variable importance of services provided and as your independent variables each of the factors influencing the importance of those services.
5.2b1b. Using the terminology of partial correlation/regression, explain how you could show the effect of each of the independent factors, holding constant the effect of the other factors.
5.2b1c. Explain why a partial correlation between each these factors and theimportance of services provided if you chose Hypothesis 1 or the (holding constant the other factors) is a better measure of the relationship than simple correlations between each factor and the importance of services provided.
Or
5.2b2a. Based on Hypothesis 2, write out a regression equation using as your dependent variable importance of services provided and your independent variable type of services. Then write a multiple regression equation using as your dependent variable importance of services provided and as your independent variables type of products and type of service. Then explain how the information in these equations can be used to determine if type of product moderates of the relationship between type of service and importance of service provided.
5.2b2b. Using the terminology of partial correlation/regression, explain how you could show the effect of each of the independent factors, holding constant the effect of the other factors.
5.2b2c. Explain why a partial correlation between each these factors and the the importance of services provided if you chose Hypothesis 2 or the (holding constant the other factors) is a better measure of the relationship than simple correlations between each factor and the importance of services provided.
Case5.3 - Multiple Regression
Hise, Richard T; Gable, Myron; Kelly, J. Patrick; and McDonald, James B (1983, Summer). Factors Affecting the Performance of Individual Chain Store Units: An Empirical Analysis. Journal of Retailing, 59(2). 22-39.
Available on 2011 January 19 in EBSCOhost via the Trident Univeristy / Touro College eLibrary.
Hise, Gable, Kelly,,and McDonald (1983) tried to predict Retail Store Performance based on 18 Variables.
Hise, Gable, Kelly,,and McDonald (1983) measured Performance in terms of
Sales Volume
Contribution Income, and
Return on Assets
For each of these measures of performance, Hise, Gable, Kelly,,and McDonald (1983) ran StepWise regressions.
5.3. What is the research design?
5.3a. What is a stepwise regression?
5.3b. As is shown in Table 2 in Hise, Gable, Kelly,,and McDonald (1983), explain why a predictor (for example, Hours Worked Per Week, in this case), which, on its own, is not a significant predictor (for example of Sales Volume, in this case), remains as a step in the Multiple Regression.
5.3c. What is multi-colinearity?
5.3d. Why is it important to estimate muli-colinearity among independent variables in a Multiple Regression?
CASE5.4 - Multiple Regression
Tharenou, Phyllis (2001, December). The relationship of training motivation to participation in training and development, Journal of Occupational and Organizational Psychology. 74 (5). 599-622.
Available on 2011 January 19 in EBSCOhost via the Trident Univeristy / Touro College eLibrary.
Tharenou (2001) assessed how training motivation, in terms of the expectation of gaining valued outcomes and motivation to learn, explains participation in training and development. Direct, mediator, and moderated explanations were tested. Survey data were gathered at Time 1 and a year later at Time 2, providing a longitudinal sample of 1705 Australians. Multiple regression analyses show that, the higher the training motivation, the more employees participated in training and development in the next 12 months, as they also did from higher supervisor support. Training motivation did not mediate the effects of the work environment on participation but moderated the prediction by employer support. Employer support predicted participation in training and development in the next 12 months more for employees with higher than lower training motivation.
5.4. What is the research design?
5.4a. Why did the author choose multiple regression analysis to show the relationship between a set of independent variables and a single dependent variable?
5.4b. Given the multiple regression analysis the author conducted, how can we interpret the significance of the various coefficients the author shows to be significant or not to be significant.
EG - on page 612, the author's direct explanation of her findings is.
As can be seen from Step 2 (Table 4; all p