Almost all marketing research projects are interested in information about a large population such as all families with children at home or all retail grocery stores. As it is impractical to collect data from all members of such large populations, a sample is selected. Various types of samples are possible, but they can be classified into two general categories – non-probability and probability.
Problems in Achieving the Scientific Method: The source of error resulting from the type of sample used comes from the possible selection of a sample that is not representative of the population with regard to the topic under study. The result will be a lack of reliability.
Minimizing Potential Sources of Error:
The first task in sampling is to define carefully just what groups of people, stores, or other units are to be sampled. For example, if the study calls for collecting data from appliance dealers. It is necessary to define what is meant by an appliance dealer (are discount stores appliances to be included?) It is also necessary to define the precise geographical area of interest (e.g. the metropolitan Chicago area).
The researcher must then decide on the type of sample which is to be selected. Probability methods use a procedure that ensures that each member in the group from which the sample is to be drawn has a known probability of being chosen.
Non-probability samples can be selected in a variety of ways, but none of them assures that each member of the population has a known probability of being selected for the sample. Because of this, non-probability sampling techniques have a greater chance of bias than probability sampling techniques.
Determining Sample Size:
The researcher must also decide how large a sample to select. Marketing research samples vary from fewer than 10 to several thousand. The researcher must consider the problem at hand, the budget, and the accuracy needed in the data before the question of sample size can be answered.
Problems in Achieving the Scientific Method:
Errors resulting from sample size are likely to be larger when small rather than large samples are used, for small samples have lower reliability.
Example: If 18 percent of a sample of Chicago’s households reports that they regularly view a certain television program, a manager can ask: How close is this sample estimate to the true percentage that exists among all Chicago household? The sample size will be a factor in determining the answer to this question. If a sample of 1,000 households was used to gather the information, the manager can be confident that the sample was large enough to provide an estimate within one or two percentage points of the actual percentage of all Chicago households regularly viewing the program. If only 50 households were sampled, and 9 households (18 percent) indicated they were regular viewers, the actual percentage of households viewing the program might be as low as 10 or as high as 30.
Minimizing Potential Sources of Error:
The magnitude of potential error from sample size can be calculated using the theory of sampling statistics if a probability sample is used. That theory can help researchers determine what sample size is needed for a given degree of accuracy. The accuracy needed in the study and the costs of using various size samples will determine the choice and, therefore, the reliability of the results.
Does the Sample Type Used by Presidential Election Polls affect the results of those polls?
The results of different presidential election polls differed noticeably. Did the different polls use different procedures to select respondents and, if so, did those different procedures have an effect on the polling results?
The Harris poll sample seems to consistently include more Democratic leaning respondents, while the NBC sample seems weighted toward the Republicans, contends Seymour Martin Lipset, a Stanford University political scientist and co-editor of Public Opinion magazine.
Both polls seek representative samples not of the general population but of the voting population, the Stanford professor says. But the definition of just what is representative of the voting population involves intuition as well as science. The Harris poll seems to give greater weight to the views of low income, minority voters who lean toward the Democrats, he says.