Lauer and Asher claim that surveys are useful when a researcher needs to determine something that can be assumed (with high confidence) as representative of a large group. Whereas case studies examine in detail an isolated group in order to name and describe variables for that group, surveys that are “statistically designed…can be considered representative of the entire population” (54). The greatest benefit with surveys, if done correctly, is that a researcher can gain valuable, generalizeable information that can be applied to a large population without the time, effort, and cost of meeting with and interviewing everyone in that population. The question, then, that I would raise from this information that Lauer and Asher present is this: is the difference that separates survey/sampling research from case studies the fact that surveys have the potential to be generalized if the confidence limits are strict enough whereas case studies are never generalizeable?
The selection of subjects for surveys is both important and difficult. Subjects are chosen by random sampling but with a group large enough to represent the population in order to have strict confidence limits. I can imagine this is very difficult. One of the greatest hindrances to selecting an effectively representative subject pool is the fact that researchers must rely heavily on the subjects’ willingness to take the time to fill out a survey. Without incentive—which often includes cost or “extra credit”—few people feel motivated to fill out a survey. Even with incentive it can at times be challenging. This is obvious when we look at Rainey et al’s “Core Competencies” article where they invited 587 people to participate and only 47 initially responded. This poor showing makes the information less generalizable, thus requiring the authors of the article to apologize for their data by stating that “the data can be assumed to be suggestive for the profession, if not representative because of the small, non-probabilistic sample.” Of course, this doesn’t discredit what they learned from their study, and valuable information was still taken. However, this then turns more to a case study of willing managers rather than a sampling that can be representative—which is the purpose of survey and sampling, is it not?
Three types of data can typically be counted: nominal, interval, and rank order. These data can be viewed easily and effectively by using the K data matrix, which aligns subjects with the variables. Nominal data is simply counting things like, as L&A show, the number of comma faults in a composition. Interval data show things like scores that have numerical intervals between them—test grades and the like. Rank order data assigns ranks from 1 to ‘n.’