A sample size is a small percentage of a population that is used for statistical analysis. For example, when figuring out how many people would vote for a certain person in an election, it isn't possible (either financially or logistically) to ask every person in the United States about their voting preference. Instead, a small sample of the population is taken. The sample size could equal a few hundred, or it could equal a few thousand. It all depends on what characteristics you want that population sample to have, and how accurate you want your results to be.
Low Sampling Error
Every time you poll a sample of a population (as opposed to asking everyone), you're going to get some statistics that are a little different from the "true" statistics. This is called sampling error, and is often expressed as percentage points. For example, a poll might be plus or minus "ten points." In other words, if a pollster finds that 55 percent of people will vote for a certain candidate, plus or minus ten points, they are really saying that somewhere between 45 and 65 percent will vote for that candidate. A good sample will have a low sampling error (a point or two).
High Confidence Level
The confidence level is based on the theory that the more often you sample a population, the more the data resembles a bell curve. Confidence levels are expressed as a percentage, such as a "90 percent confidence level." The higher the confidence level, the more sure a researcher is that his data looks like a bell curve: a 99 percent confidence level is desirable and likely to have better results than a 90 percent (or lower) confidence level.
Degree of Variability
The degree of variability refers to how diverse a population is. For example, a poll of all political parties about health care is likely to result a more widespread variation in responses than a simple poll of a single party. The higher the stated proportion, the greater the level of variability, with .5 being the highest (and possibly, least desirable) value. For smaller samples, you would want to see a low degree of variability (for example, .2).
About the Author
Stephanie Ellen teaches mathematics and statistics at the university and college level. She coauthored a statistics textbook published by Houghton-Mifflin. She has been writing professionally since 2008. Ellen holds a Bachelor of Science in health science from State University New York, a master's degree in math education from Jacksonville University and a Master of Arts in creative writing from National University.
Photo Credits
Africa statistics image by michele goglio from Fotolia.com