Saturday, February 21, 2009

Week 7: Sampling & Surveys

Appropriate Purposes for Surveys
Surveys can appropriately be used to make large descriptive tasks possible with a minimum of cost and effort. (Sampling surveys describe a large group (population) in terms of a sample (smaller part of group). It concentrates on a few variables of small groups, which can be representative of a larger group.) It allows researchers to obtain readily observed or recalled behavior, and can be related to several major demographic characteristics. It helps reduce large populations to a manageable size because of sampling procedures and provides a valuable means of obtaining representative descriptive information (not cause/effect).

Surveys make the research effort more reasonable. Researchers should look into the question of feasibility (# of units) from which they can collect good data and also adequately analyze it.

Subject Selection
The simplest and best strategy for subject selection is random selection, in which the number of the population selected for study is put into alphabetical order, and then selected randomly by hand (long and involved), or ideally by calculator or computer (using a random number function). Data is collected using questionnaires (scored, and open ended questions). Other types of sampling include systematic random sampling (useful when the population to be studied is already organized in a sequence in the data); quota sampling (helpful when a researcher knows the % of specific features of the population); stratified samples (when some parts of the population are of more immediate interest); cluster samples (when the researcher wishes to study individual units within a large population. This type of sampling should be avoided unless the researcher is using or is a strong statistician).

Collecting and Analyzing Data
Questionnaires (scored and open ended questions) and surveys can be used. Prime consideration should be given to the capability of eliciting a high response rate. Response rates are part of analyzing data, and depend on following sampling size, and paying attention to the nonresponse to questionnaires (remember to chase the data - make phone calls, and try to make two or three attempts to obtain the data). 3 types of data can be collected in a survey: nominal (simple counting/#'s and %'s); interval (usually comes from test scores w/large #'s of items); rank (subjects placed in hierarchical order, which is assumed to be equal order) (p.70-74 L&A).

Possible Generalizations
This is descriptive research, so one must be careful drawing conclusions from the results, as it is often difficult to make cause/effect statements. These generalizations should be made only to the population from which the sample was drawn (p.78 L&A).

(Material taken from Lauer & Asher)

1 comment:

  1. Wendy, did you find it difficult discerning how far was too far concerning cause/effect generalizations? In the other readings, I found myself throwing up a finger--AHA!--thinking I've found a researcher trespassing on quantitative ground, i.e., drawing a cause/effect conclusion. After throwing up my finger multiple times, I started wondering if my trigger was too sensitive: I was finding cause/effect statements scattered throughout the endings of most readings. The ground leading up to them seems slippery as wet tile, on a slope, with a strong wind at your back. I hope we are able to discuss how exactly you can creep up to a cause/effect conclusion without actually writing one. That's where the persuasive power seems to lie: to do everything you can with your research to indicate there is some kind of correlation, some kind of undeniable relationship, without stepping over that thin line separating causes from their effects. But the descriptive mode keeps lurching, wanting to leap, wanting to get to the other side. Or am I, again, just a little too sensitive about this?