Sampling Issues

What population is the survey investigating? While several restrictions were placed on the demographics of the respondees, i.e. only Christian, the voluntary nature of the eliciting process, along with the undefined nature of the group to which the survey was advertised, makes delineation of a larger population of interest intractable. As with the previous coin example, to be able to make inferences about a larger group, we must understand the relationship between the likelihood of making an observation and the nature of the observation itself.

For example, if we wish to make inferences about the opinions of students attending SPU, we need to come up with a sampling strategy through which we can understand the relationship between sampling likelihood and sample outcome. To avoid making extra work, one often would like a sampling scheme such that  the parameter estimates from the samples will be an unbiased estimate of the population parameters. One such simple strategy is that of simple random sampling. In this situation, there is an equal likelihood of sampling from any given student at SPU. There are, of course, a range of sampling strategies for a variety of situations.

Obtaining a representative sample may be difficult; if one were to go to an institution and collect surveys from volunteers, one could run into a problem in which people who volunteer and complete a survey do not have opinions which are representative of the larger population. And one suspects that such could be the case for the topics of Christian faith and evolution. (For example, people with strong opinions may be more likely to volunteer to take a survey than those with weak or conflicted opinions.) If, however, one could, with uniform randomness, select a pool of students at an institution and elicit their opinions, with a very low rate of refusal, then, one would be in a position to make inferences about the larger group.

Email link | Printer-friendly | Feedback | Contributed by: David Caccia