Home » Business

Basics of Choosing a Survey Population

Sep 13, 2008
Most surveys are conducted only after the goal of the data has been determined. Depending upon the objective of the data, a target group is identified. When that target group is large, surveying the entire group is often impossible. A representative sample of that group must be chosen. Typically, a random selection process is used to build a survey population (i.e. the people who will participate in the survey). While most people imagine that choosing a survey population and having confidence in the survey results is straightforward, it is not. Below, we'll explain what is involved when choosing a survey population. You'll also learn why survey results can be misleading.

Random Sampling, Subgroup Selection, Or Both?

Random sampling is an effective way to measure the perspective of a group of people that is too large to interview. For example, surveying Americans about a political candidate would be impossible due to the number of Americans. In this case, a random sample is chosen as the survey population. Conclusions are drawn about Americans' view of the political candidate based upon the results from interviewing the smaller sample.

Often, you'll want to divide your random sample into separate groups of people. This is done when the perspective of the groups are expected to vary widely. For example, assume that you have chosen a sample population to survey about the aforementioned political candidate. Your population contains both men and women. Because these two groups often perceive political candidates differently, subgroup selection and interviewing may be worthwhile.

It's important to note that doing this can require a significant amount of additional time. If the effort and time involved is too great, you can perform what it called a stratified random sampling of the disparate groups (a topic that is beyond the scope of this article).

How Accurate Are Your Results?

The accuracy (and thus, reliability) of survey data is often misunderstood. While most people intuitively understand that the survey results of a random sampling will contain some allowance of error, they may not know how the error rate and confidence level is derived. Exacerbating the issue is the fact that these numbers are often unreported. Let's return to our previous example to illustrate this concept.

Assume once again that you're surveying Americans about the political candidate. You obviously cannot survey every American. So, you choose a sample population to interview. The level of accuracy in your results will reflect how large a percentage your sample population is in comparison to your larger group (i.e. Americans). The larger your sample, the more accurate you can expect your results will be. Conversely, the higher error rate you're willing to assume in your data, the smaller your sample population must be.

Making Corrections When Needed

There will be times when a portion of your random sample population fails to complete your survey. This can present a problem. When you choose the number of people for your population, you do so based upon the error rate you're willing to assume and the level of confidence you would like to have in your data. When some members of your population fail to respond to your survey, your original assumptions of error and confidence are no longer reliable. To circumvent this problem, estimate a percentage of non-response and build your population based upon your estimate.

Choosing your survey population is a matter of making reasonable assumptions and having a clear objective regarding the purpose of your data. Just keep in mind that assumptions are often wrong. Learn to make corrections along the way.
About the Author
SurveyGizmo is a leading provider of online survey software, check them out on the web for more great ways to use surveys to enhance your business.
Rating:
Please Rate:
(Average: Not rated)
Views: 127
Print Email Report Share
Article Categories