Part 1
A marketing survey leads to a collection of different types of data about a particular market or consumer group. The data collected can be from various elements of a market (e.g., a sample of car owners or night shoppers). If we are going to base our decisions on the results achieved through market research then, the research has to be highly valid.
There is an often quoted statement “There are three types of lies---lies, damn lies, and statistics” by Benjamin Disraeli. This statement is true if the statistical procedures are not applied prudently and authentically. There are two main sources of misuse of statistics: 1. a deliberate plan on the part of a dishonest researcher 2. Unintentional errors occurring in the process of research. Both are very serious and should be tackled with scientific rigor.
Here is a case in our hand, in which a market survey has progressed to the stage of data collection. About 1947 filled in questionnaires have been returned by the target sample, out of which 310 questionnaires are partially completed. (Missing data). The acceptable numbers of survey samples are now reduced to 1637. Though the sample size of 1600 is large enough, if we increase the sample size, the extent of sampling error (standard error) can be reduced. Sampling error is the difference between a statistic and the values we expect over many repetitions of sampling. The principle followed to reduce sampling error is, the larger the sample size, the better it is. Large sample size will take the statistic nearer to the population parameter. Hence, the first action I would explore is whether I can complete the partially filled questionnaire. I would ask the student to contact the consumers and resend the questionnaires and persuade them to complete the survey. If we can add more completed surveys, the sampling error is curtailed.
There is another set of possible errors that can affect the research results, they are non sampling errors. Non sampling errors are more serious, which occur due to faults made in the collection of data or improper selection of data. Three kinds of non-sampling errors can occur. 1. Errors in data acquisition. 2. Nonresponse errors. 3. Selection bias. In order to reduce the gravity of these errors, I would direct the student to contact the respondents, gain their cooperation and conduct an actual interview based on a questionnaire sent. With this process two advantages are achieved, one we are establishing reliability of our questionnaire and another we are benchmarking the quality and the trends in responses. If the short and limited interview is in line with the received data, we can create a benchmark and then compare all the received data with the benchmark. If any inconsistency is spotted in the completed questionnaires, we can review with the respondent. With this process, we can make our data robust for data analysis and research reporting. Loosveldt, G., Carton, A., & Billet . J. (2004) has proposed a similar procedure. Finally, I would reject those returned questionnaires that are incomplete, that are not authentic or that has inconsistent responses.
Part 2
When we set out to do market research, we are inevitably trying to test a hypothesis. Quantitative market research commonly has some quandary, issue or problem that needs investigation.. The purpose of hypothesis testing is to find ways to test effectively our assumptions. As you know there are two types of hypothesis, Null hypothesis and Alternate hypothesis. At the time of data analysis, when we are close enough to accept or reject a null hypothesis,a rigour is required to set the significance level for our research.
Hypothesis testing involves specifying a significance level of our research. Our research can have errors in hypothesis testing, and there are two types of errors. Type I and Type II. A type I error is rejecting the null hypothesis when it is true, and the type II error is accepting the null hypothesis when it is false.
The scientific reason for distinguishing type I and type II errors are that, depending on the research situation, it is important to understand which type is more important to avoid. Based on this information we need to orient the research methodology, data analysis and significant level.
The significance level is about probability. What is the probability that people are buying due to attractive packaging? And at what level is the probability? We may have two types of packaging; one is very attractive, and another one is ordinary. If we have a number of retail outlets, where we keep either attractive package or ordinary package in each of the outlets, then what are the chances that the mean sales of the packages are identical . If the mean is different, is it because of chance factor or due to packaging difference? For answering such a question, we need to fix the level of significance. If I want to be sure that the packaging difference ‘really’ has an influence on buying behavior, I will choose 0.01 (a 1% chance or less; 1 in 100 chance or less) level of significance. This level of significance used in this situation because we have to reject the null hypothesis. Iyer, K. S., Srivastava, P., & Rawwas, M.A. (2014) has done an excellent study on supply chain management, with several hypothesis and applying statistics at both significance levels .01 and .05 levels.
Part 3
Science and scientific management is all about finding relationships between phenomena. Marketing research too is a science because it attempts to understand the consumer beahviour. Regression has very important place in market research and it builds on the values of correlation, establishing relationships among variables to make predictions. Regression is a tool to predict the value of a dependent variable from the a single or a set of independent variable.
Though regression is a very basic and important statistical tool, there are studies which do not use regression as a statistical tool. The statement, “Regression is such a basic technique that it should always be used in analyzing data”, is superfluous. The analysis of data involves descriptive statistics such as mean , mode median, standard devisation etc, and inferential statistics which includes ANOVA and regression analysis. There is a point of view that both ANOVA and regression are same, with different names. Armstrong, J. S. (2012) says “ Regression analysis can play an important role when analyzing non-experimental data”, but do not recommend if you are searching for causal relationships. Olsen, C. (1996) gives a good account of the difference between ANOVA and regression, and there are market research studies which only uses ANOVA. From this it can be inferred , regression need not be used always.
Part 4
The pupose of reports are to examine available and potential solutions to a problem, situation, or issue and reach conclusions about those problem or issue. The usual structure of a report has an introduction, discussion, conclusion and recommendations. But there are no standards how long a report shoul be. The report can be as long as a single page or can have even thousand pages . The kind of report, the audience who would read it and the purpose of report determines the size of the report.
If a report can communicate what it intends to in least possible way, then a report can be concise and yet cover the relevant details. The statement “Writing a report that is concise and yet complete is virtually impossible as these two objectives are conflicting.” is not true. From your day to day experience, several examples of single page reports can be drawn, which are short yet providing complete informations , for example report of blood analysis, academic scores etc.
REFERENCES
Armstrong. J. S. (2012) Illusions in Regression Analysis, International Journal of Forecasting, 2012 (3), 689-694.
Bailey, K. D. (1978). Methods of social research. New York: Free Press.
Berg, B. L. (1995). Qualitative research methods for the social sciences. Boston: Allyn and Bacon.
Bowden, J. (1997). Writing a report: A step-by-step guide to effective report writing (4th Ed.). Plymouth, England: How to Books.
Fink, A. (1985). How to conduct surveys: A step-by-step guide. Beverly Hills: Sage Publications.
Groves, R. (1989) Survey Errors and Survey Costs. New York: John Wiley.
Iyer, K. S., Srivastava, P., & Rawwas, M. A. (2014). Aligning Supply Chain Relational Strategy with the Market Environment: Implications for Operational Performance. Journal of Marketing Theory & Practice, 22(1), 53-72.
Kalsbeek, B. (1995). How to Plan a Survey. www.parkdatabase.org//1999_ASA-How-to-plan-a-survey_ASA.pdf, retrieved 18.02.2014
Kalsbeek, B. (1995). How to Collect Survey Data. http://www.parkdatabase.org/files/documents/1999_ASA-How-to-collect-survey-data_ASA.pdf, retrieved 18.02.2014
Loosveldt, G., Carton, A., & Billiet, J. (2004) Assessment of Survey Data Quality: A Pragmatic approach focused on Interviewer tasks. International Journal of Market Research, 46(1), 65-82.
Oslen, C (1996). Difference between ANOVA and Regression, http://www.cscu.cornell.edu/news/statnews/stnews13.pdf retrieved 20.02.2014